Cloud Computing

Module 2

1
 Introduction to Cloud Computing

• Definition by National Institute of Science and

Technology (NIST) USA:
Cloud computing is a model for enabling ubiquitous,
convenient, on-demand network access to a shared
pool of configurable computing resources (e.g.,
networks, servers, storage, applications, and
services) that can be rapidly provisioned and released
with minimal management effort or service provider
interaction.

2
Introduction to Cloud Computing
• Essential Characteristics
according to NIST
definition:
• On-demand self-service
• Broad network access
• Resource pooling
• Rapid elasticity
• Measured service
Introduction to Cloud Computing
• On-demand self-service: A
consumer can unilaterally
provision computing
capabilities, such as server
time and network
storage, as needed
automatically without
requiring human
interaction with each
service provider.
Introduction to Cloud Computing
• Resource pooling: The
provider’s computing
resources are pooled to
serve multiple consumers
according to consumer
demand. The customer
generally has no control
or knowledge over the
exact location of the
provided resources (e.g.,
country, state, or
datacenter). Examples of
resources include storage,
processing, memory, and
network bandwidth.
Introduction to Cloud Computing
• Resource pooling: The
provider’s computing
resources are pooled to
serve multiple consumers
according to consumer
demand. The customer
generally has no control
or knowledge over the
exact location of the
provided resources (e.g.,
country, state, or
datacenter). Examples of
resources include storage,
processing, memory, and
network bandwidth.
Introduction to Cloud Computing
• Measured service: Cloud
systems automatically
control and optimize
resource use by
leveraging a metering
capability. Typically this is
done on a pay-per-use or
charge-per-use basis.
Cloud Computing
Module 3

8
Brief History of Cloud Computing
• Computer Scientist John
McCarthy is attributed
with delivering the idea
that computations will be
provisioned as utilities in
future. This idea was
presented in 1961.
Brief History of Cloud Computing
• Computer Scientist John
McCarthy is attributed
with delivering the idea
that computations will be
provisioned as utilities in
future. This idea was
presented in 1961.
Brief History of Cloud Computing
• In 1960s and 1970s, the
mainframes (giant
powerful computers)
were leased out by the
manufacturers.
• The idea of grid
computing emerged in
1990s to use the
processing power of
networked PCs for
scientific calculations
during idle times.
Brief History of Cloud Computing
• In 1990s, Salesforce.com
started bringing remotely
provisioned software
services to the
enterprises. Amazon Web
Services (AWS) were
launched in 2002.
Brief History of Cloud Computing
• In 2006, the term “cloud
computing” emerged that
enabled organizations to
“lease” the computing
capacity and processing
power from cloud
providers.
Brief History of Cloud Computing
• In 2006, the term “cloud
computing” emerged that
enabled organizations to
“lease” the computing
capacity and processing
power from cloud
providers.
Cloud Computing
Module 4

15
Overview of Cluster Computing
• A computer cluster is a
collection of interconnected
stand-alone computers
which cooperate together
to work as a single resource
pool of computing
resources.
• Clusters became popular in
1990s when mainframes
and traditional
supercomputers were
becoming less cost-
effective for high
performance computing
(HPC).
Overview of Cluster Computing
• In 2010, out of top 500
supercomputers. 85% were
computer clusters built with
homogeneous nodes.
• Cluster computing has laid
the foundation of modern
day super computers,
computational grids and
cloud computing.
Overview of Cluster Computing
• Important Benefits of
Cluster Computing:
• Scalability
• High availability and fault
tolerance
• Use of commodity
computers
Overview of Cluster Computing

Sequential and Parallel Parallel Programming Environment

Applications
Cluster Middleware Ensuring High Availability and Single System Image

PC/Workstatio PC/Workstatio PC/Workstatio

n Communication n Communication n Communication
Software Software Software

Network Interface Network Interface Network Interface

hardware hardware hardware

Cluster Interconnect Network Switch

Cloud Computing
Module 5

20
Overview of Grid Computing
• The grid is an integrated
computing infrastructure
for bringing together
computers to create a large
collection of compute,
storage, and network
resources.
Overview of Grid Computing
• Grid is used to solve large-
scale computation
problems or to enable fast
information retrieval by
registered users or user
groups.
• Computers include PCs,
workstations, server
clusters, supercomputers,
laptops, notebooks, mobile
computers, PDAs, etc.
Overview of Grid Computing
• Building virtual grid through
CPU scavenging: Creating a grid
by using unutilized CPU cycles
in a network of computers at
night or periods of inactivity.
• This is done on voluntary basis.
The grid hosts donate some
RAM, disk space and network
bandwidth as well.
• The most famous example is
the SETI@Home which applied
over 3 million computers to
achieve 23.37 TFlpos as of Sept.
2001.
 Overview of Grid Computing
• Application: The top layer consisting of user applications
to be run on grid.
• Collective Services: Focus on interaction among the
resources. implements functions such as resource
discovery, scheduling, brokering etc.
 Overview of Grid Computing
• Resource service: Deals with the aggregated computing
resources (software and hardware) available for user
applications in collective operations.
• Connectivity Layer: Provides the core networking among
the computational resources of fabric layer through
physical or virtual networking.
• Grid fabric: Consists of all the computational resources
such as storage systems, catalogs, network resources,
servers and their network connections.
 Overview of Grid Computing
• Grid Middle ware Layered Architecture (deployed on
participant computers)

Application

Collective services

Resource services

Connectivity services

Grid fabric
Cloud Computing
Module 6

27
Difference between Cluster, Grid and Cloud
Computing

• The purpose of Grid Computing

is to solve large scale
computational problems.
• Just like Clusters, except that
• The Grids make use of
computational resources are
spread across the nation or
the globe.
• These computational
resources are owned by
different organizations and
are shared (as grid
resources) by multiple users.
Difference between Cluster, Grid and Cloud
Computing

• Grids heavily depend upon

WAN/LAN resources.
• Virtual Supercomputer term is
derived from Grid Computing
whereby multiple computers
collaborate together over
network to create an illusion
of a single big computer.
Difference between Cluster, Grid and Cloud
Computing

• Grids heavily depend upon

WAN/LAN resources.
• Virtual Supercomputer term is
derived from Grid Computing
whereby multiple computers
collaborate together over
network to create an illusion
of a single big computer.
Difference between Cluster, Grid and Cloud
Computing

• Grid Computing:
• As compared to cloud:
• The resources do not join or
leave the grid dynamically.
• Majority of the resources are
not provisioned from data
centers.
• Several organizations may
unite to form a grid in the
shape of a virtual organization
(VO). For example multiple
hospitals and research centers
may collaborate in a VO to find
a cure for cancer.
Cloud Computing
Module 7

32
Cloud Service Models
• Service Models according
to NIST definition:
• Software as a Service
(SaaS)
• Platform as a Service
(PaaS)
• Infrastructure as a
Service (IaaS)
Cloud Service Models
• Software as a Service
(SaaS):
• The capability provided
to the consumer is to
use the provider’s
applications running on
a cloud infrastructure. A
cloud infrastructure is
the collection of
hardware and software
that enables the five
essential characteristics
of cloud computing.
Cloud Service Models
• Platform as a Service
(PaaS):
• The capability provided
to the consumer is to
deploy onto the cloud
infrastructure consumer-
created or acquired
applications created
using programming
languages, libraries,
services, and tools
supported by the
provider.
Cloud Service Models
• Infrastructure as a Service
(IaaS):
• The capability provided
to the consumer is to
provision processing,
storage, networks, and
other fundamental
computing resources
where the consumer is
able to deploy and run
arbitrary software,
which can include
operating systems and
applications.
Cloud Service Models
• Infrastructure as a Service
(IaaS):
• The consumer does not
manage or control the
underlying cloud
infrastructure but has
control over operating
systems, storage, and
deployed applications;
and possibly limited
control of selected
networking components
(e.g., host firewalls).
Cloud Deployment Models
• Cloud Deployment
Models according to NIST
definition:
• Private cloud
• Community cloud
• Public cloud
• Hybrid cloud
Cloud Deployment Models
• Private cloud: The cloud
infrastructure is
provisioned for exclusive
use by a single
organization comprising
multiple consumers (e.g.,
business units). It may be
owned, managed, and
operated by the
organization, a third
party, or some
combination of them, and
it may exist on or off
premises.
Cloud Deployment Models
• Community cloud: The cloud
infrastructure is provisioned
for exclusive use by a specific
community of consumers from
organizations that have shared
concerns (e.g., mission,
security requirements, policy,
and compliance
considerations). It may be
owned, managed, and
operated by one or more of
the organizations in the
community, a third party, or
some combination of them,
and it may exist on or off
premises.
Cloud Deployment Models
• Public cloud: The cloud
infrastructure is
provisioned for open use
by the general public. It
may be owned, managed,
and operated by a
business, academic, or
government organization,
or some combination of
them. It exists on the
premises of the cloud
provider.
Cloud Deployment Models
• Hybrid cloud: The cloud
infrastructure is a
composition of two or
more distinct cloud
infrastructures (private,
community, or public).
Cloud Computing
Module 8

43
Basics of Computers
• Mainframe:
• A mainframe is a large,
expensive, powerful
server that can handle
hundreds or thousands
of connected
users/servers
simultaneously. For
example a single
mainframe server of
IBM’s Z series can
provide the equivalent
computing throughput
of at least 500 servers.
Basics of Computers
• In 1960s and 1970s, the
mainframes were leased
out by the manufacturers
rather than sold because of
enormous cost of
ownership.
Basics of Computers
• Mainframe leasing model:
• The customers were
charged on monthly basis
for the use of hardware
such as CPU, memory and
peripheral devices.
• The software (compilers,
editors etc.) usage was
charged for the time of
usage.
Basics of Computers
• The mainframe leasers
used to develop
customized software
exclusively for a client
organization and charged
for it.
• The client was also
charged for the
maintenance of those
customized software.
• This model still exists in
the form of cloud
computing.
Basics of Computers
• Server:
• A server is a computer
which provides services
to other computers
and/or devices connected
to it. Services provided by
a server include the
controlled access to
hardware and software
resources and storage.
• A server can support
hundreds and thousands
of simultaneous users.
Basics of Computers
• Servers are available in a variety
of sizes and types:
• Web server: stores websites
and web apps and provides
them on your desktops and
mobiles through web
browsers.
• Domain Name Server (DNS):
Stores domain names and the
corresponding IP addresses.
• Database server: Hosts
database and provides access
to data and provides data
manipulation functionality.
Basics of Computers
• Desktop
• A desktop is a computer
which is designed to remain
in a stationary position. It is
used as a personal computer.
• Performs all of its activities
such as
• Input
• Processing
• Output
• Storage
by itself.
• Intended to be used by one
person at a time.
Cloud Computing
Module 9

51
Basics of Data Communication
• Data Communication:
• Exchange of data over
some transmission
medium between two
devices.
Basics of Data Communication
• The following factors are
essential for data
communication:
• Data must be delivered to
correct destination.
• Data must not be altered
during transmission.
• There must be timely
delivery of the data.
• Data must not be altered
during transmission.
• There must be timely
delivery of the data.
Basics of Data Communication
• There must not be
uneven delay among the
packet arrival time during
audio or video
transmission.
Basics of Data Communication
• Components:
• Message: The data to be
sent. Can be text,
numbers, pictures,
audio and video.
• Sender
• Receiver
Basics of Data Communication
• Transmission medium:
The physical path through
which a message travels
from sender to receiver.
• Protocol: The set of
agreed-upon
communication-rules
between sender and
receiver devices. Two
devices can be connected
but not communicating
without a protocol.
Basics of Data Communication
• Data Representation:
• Text: Represented by bit
pattern called code e.g.;
Unicode and American
Standard Code for
Information Interchange
(ASCII).
• Numbers: Directly
converted binary of the
number. ASCII is not
used to represent
numbers.
Basics of Data Communication
• Images: Sent as binary
patterns. Image is
represented by a matrix
of pixels. Pixel is a small
dot. Each pixel is
assigned a bit pattern
on the basis of color.
• Audio: A continuous
stream of data.
Different from text,
numbers and images.
• Video: Can be a
continuous stream or a
sequence of image
combinations.
Basics of Data Communication
• Images: Sent as binary
patterns. Image is
represented by a matrix
of pixels. Pixel is a small
dot. Each pixel is
assigned a bit pattern
on the basis of color.
• Audio: A continuous
stream of data.
Different from text,
numbers and images.
• Video: Can be a
continuous stream or a
sequence of image
combinations.
Cloud Computing
Module 10

60
Data Flow
• Data Flow:
• Simplex:
• Unidirectional
communication in
which either one of
the sender or
receiver device can
transmit. For
example: key board,
monitor etc.
Data Flow
• Data Flow:
• Simplex:
• Half Duplex: Both
devices can
communicate but
one at a time. The
entire capacity of the
transmission
medium is available
to the transmitting
device. For example:
Walkie-talkies.
Data Flow
• Data Flow:
• Simplex:
• Half Duplex: Both
devices can
communicate but
one at a time. The
entire capacity of the
transmission
medium is available
to the transmitting
device. For example:
Walkie-talkies.
Data Flow
• Full Duplex:
• Both devices can
send and receive at
the same time. The
transmission
medium should
provide separate
paths (channels) for
the transmission of
each device. For
example telephone
conversation is full
duplex.
Data Flow
• Full Duplex:
• Both devices can
send and receive at
the same time. The
transmission
medium should
provide separate
paths (channels) for
the transmission of
each device. For
example telephone
conversation is full
duplex.
Cloud Computing
Module 11

66
Basics of Computer Networking
• Computer networking was
conceived in 1960s soon
after the invention of
computers.
• A network is a collection of
computers and devices
connected together
through transmission
media.
Basics of Computer Networking
• Devices:
• Hosts: Large computers,
desktops, laptops,
cellular phone or security
system.
Basics of Computer Networking
• Connecting devices:
• Router: A device which
connects the network
with other networks.
• Switch: A device which
connects devices within
the network.
• Modem: A device which
changes the form of data
(modulates-
demodulates).
Basics of Computer Networking
• Network Criteria:
• Performance: It is often
evaluated by two metrics:
• throughput (bulk of data
transmitted in unit of
time) and delay.
• Increasing the
throughput may increase
the congestion and hence
increase the network
delay.
Basics of Computer Networking
• The transit time (message
travel time) and response
time (time between
inquiry and response)
indicate the network
performance also.
Basics of Computer Networking
• Reliability: It is measured
in terms of frequency of
network failure, time to
recover from a failure and
robustness from
disasters.
• Security: Protecting data
from unauthorized access
and damage, and
implementation of
security policies and
procedures for recovery
from breaches and data
losses.
Cloud Computing
Module 12

73
Basics of Computer Networking
• Physical Structures:
• Network Connections:
Communication can
only take place if the
devices are
simultaneously
connected to the same
communication-path or
link or connection.
• A link can be dedicated
link (Point to Point) or
shared among devices
(multipoint).
Basics of Computer Networking
• Physical Topology:
• Mesh: Every device has
a dedicated point to
point link to every other
device.
• Advantage: robustness
of network from failure
of any link.
• Disadvantage: The bulk
of cabling involved.
Basics of Computer Networking
• Star: All devices are
connected to a central
device. Unlike mesh,
there is no direct traffic
between any two devices
but through the central
device such as hub.
• Advantage: Requires
only one I/O port in
each device as
compared to mesh.
• Disadvantage: If the
central device fails, the
whole network fails.
Basics of Computer Networking
• Bus: A multipoint topology
in which one long cable is
used a network backbone.
• Advantage: Ease of
installation. Requires less
cabling than mesh and star.
• Disadvantage: Difficult to
extend, signal drops along
the length of cable results
in limited number of
connections, breaking of
backbone cable isolates
the network segments and
introduces noise.
Basics of Computer Networking
• Ring: The devices are
connected in the form of
ring. Each device acts as
repeater.
• Advantages: Easy to
expand and alter the
network. Disadvantages:
• Disadvantage: Failure of a
single device can disable
the entire network,
transmitting device needs
to retain the token signal
to perform transmission
which slows down the data
rate.
Cloud Computing
Module 13

79
Network Types
• Local Area Network (LAN): It
is a privately owned network
and has a scope of an office,
building or a campus. A LAN
can even extend throughout a
company.
• Each host in a LAN has a
unique identifier or address.
• The communication packets
between any two hosts in a
LAN contain the source and
destination hosts’ addresses.
Network Types
• Local Area Network (LAN) key
features:
• Media type: wired/wireless,
twisted pair/cable/fiber,
radio, infrared
• Topologies: Bus, Star, Mesh,
Ring, Tree
• Bit rate: from 1Mbps to
1Gbps
Network Types
• Typical LANs:
• Ethernet (CSMA/CD): Carrier
Sense with Multiple Access
with Collision Detection
(retransmission after
collision detection)
• Local Talk (CSMA/CA): CSMA
with Collision Avoidance
(reserve the media before
transmission)
• Wireless LAN: IEEE 802.11,
Range: < 100 m, Speed:
2Mbps
Network Types
• Typical LANs:
• Token Ring: A token travels
around the ring, it must be
retained by the sender
computer to send a single
packet, 4,6 or 100 Mbps
• FDDI: Token ring with fiber
optic cable, 100 Mbps
• ATM: Star based, uses
switch, multiple devices can
communicate
simultaneously, 25, 45, 155,
600+ Mbps
• Unicast, Broadcast, Multicast
Network Types
• Wide Area Network (WAN): A
network that spans large
geographical area such as
town, cities, states or even
countries. Usually
interconnects multiple LANs.
• Unlike LAN which is owned
by the user organization, a
WAN is normally created
and run by communication
companies. It is leased to
the user organizations.
Network Types
• Wide Area Network (WAN)
Types
• P2P WAN: Connecting two
devices through wired or
wireless media eg;
connecting two LANs to
form a private internet or
internetwork of a company.
• Switched WAN: A network
with more than two ends. It
is a combination of several
P2P WANs connected by
switches.
Network Types
• Metropolitan Area Network
(MAN): It is a computer network
covering a large geographical
area bigger than LAN and smaller
than WAN.
• Diameter: 5 to 50 km, several
buildings or a whole city.
• MAN is not owned by a single
organization generally just like
WAN. The MAN equipment are
usually owned by a service
provider.
• MAN usually provides high
speed connectivity to allow
sharing regional resources.
Cloud Computing
Module 14

1
Switching
◦ A WAN is a switched
network in which a
switch connects two
links together to
forward data from one
network to the other.
Switching
Two common types of
switched networks are:
◦ Circuit-Switched
Network
◦ Packet-Switched
Network
Switching
• Circuit-Switched
Network:
• A dedicated physical
connection (circuit) is
present between two
devices.
Switching
• Circuit-Switched
Network:
• A switch activates or
deactivates the circuit
for each
communication
session.
• The switch does not
store the data, only
forward to the
receiver.
Switching
• Circuit-Switched
Network:
• Used extensively by
telephone companies
for voice
communications..
Switching
• Circuit-Switched
Network:
• Only useful when all
the circuits are being
utilized
simultaneously;
otherwise the
network is being
underutilized.
Switching
• Packet-Switched
Network:
• It is a WAN switching
method in which the
data consists of
packets which can be
stored and forwarded
by the switching
devices such as
routers.
Switching
• Packet-Switched
Network:
• Statistical
multiplexing is used to
enable devices to
share the packet-
switching circuits.
• More efficient than
circuit-switching.
Cloud Computing
Module 15

10
The Internet
• It is a network of
thousands of
interconnected networks.
The Internet
• Internet History:
• Internet has its roots in
the networking project
started by Pentagon’s
Advanced Research
Projects Agency (ARPA),
an agency of the U.S.
Department of Defense
in the mid of 1960s.
The Internet
• Internet History:
• The objectives of ARPA
were to develop a
network To allow the
sharing of military and
scientific information
between scientists at
different physical
locations.
The Internet
• Internet History:
• The network could
remain functional even
if partially disabled or
destroyed by a disaster
such as a nuclear attack.
• Today the ARPA
Network is known as
the Internet.
The Internet
• Accessing the Internet:
• Dial-up service
• DSL Service
• Cable TV Networks
• Fiber to the premises
• Wireless Networks
• Direct connection to
Internet by becoming
an ISP
 The Internet
• Internet today:
• World Wide Web
• Social media
• Multimedia
• Email
• Voice over IP (VoIP)
◦ File transfer (uploads
& downloads)

end
Cloud Computing
Module 16

17
TCP/IP Suite
• TCP/IP Protocol Stack:
• Transmission Control
Protocol (TCP) was
proposed in 1973 to
ensure a reliable, end-to-
end and error free
transmission control.
TCP/IP Suite
• TCP/IP Protocol Stack:
• transmission control.
• It was latter split into TCP
and Internet Protocol (IP)
layers with IP handling
the message routing and
TCP performing the error
control.
TCP/IP Suite
• Since 1981, TCP/IP is
included in the operating
systems.
TCP/IP Suite
• TCP/IP Protocol Stack:
• Consists of layers of
protocols which paved
the way for creating
today’s internet.
• These layers help in
dividing a complex task
into several smaller and
simpler tasks.
TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Application
• Transport
• Network
• Data-link
• Physical
TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Physical Layer: Deals
with transmission of bits
into signals and
transmission of signals
over the link.
TCP/IP Suite
Network
• TCP/IP Protocol Stack
Layers:
• Data-link Layer: Creates
the frames of data.
• Each frame contains the
data and is addressed
with the MAC address of
the receiving device and
also contains the MAC
address of sending
device.
TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Network Layer: Is
responsible for host to
host communication
through their IP
addresses and related
protocols.
• No control for error and
congestion is performed.
• Packets are called
datagrams.
TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Transport Layer:
Responsible for
transporting a message
from application
program running over
source host to
corresponding
application program on
destination host.
TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Transport Layer: Main
protocols are:
• Transmission Control
Protocol (TCP)
• User Datagram
Protocol (UDP)
• Works on port numbers
on corresponding hosts.
TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Transport Layer:
• TCP: Provides flow
control, congestion
control and error
control as it is a
connection oriented
protocol.
• UDP: Is light weight
and is not connection
oriented.
TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Transport Layer:
• TCP Message =
segment
• UDP Message =
datagram
 TCP/IP Suite
• TCP/IP Protocol Stack
Layers:
• Application Layer:
Consist of programs
running on two hosts
and exchanging
messages.
• Applications use these
protocols for
communication: HTTP,
FTP. SMTP
end
Cloud Computing
Module 17

31
IP Addressing
• The identifier used in
Network layer of TCP/IP
suit is the address of the
internet connection of
receiver and sender
devices.
• IPv4 is a 32 bit universally
unique address while IPv6
is the 128 bit universally
unique address.

32
IP Addressing
• Total IPv4 addresses = 232
• Total IPv6 addresses = 2128

33
IP Addressing
• The address is in fact of
the connection and may
change when the device is
moved to another
network.
• A device can have two IP
addresses if it has two
connections with the
internet.

34
IP Addressing
• IP address is usually
represented by dotted
decimal numbers. For
example:
• IP v4 address: 193.63.82.10

35
IP Addressing
IP Addressing:
◦ The IP addresses are
allocated by the Internet
Corporation for Assigned
Names and Numbers
(ICANN) to ISPs and large
organizations.
◦ Smaller organizations can
get IP addresses from
ISPs.

36
IP Addressing
IPv4 Addressing:
• The IP address consists of
a prefix part (the Network
ID) and postfix part (the
Host ID or the Subnet).

37
IP Addressing
• Classification of IPv4
addresses:
• Class A: 8 bits for
Network ID
• Total networks 27
• Network id starts with
‘0’ binary
• First byte: 0 to 127
◦

38
IP Addressing
• Classification of IPv4
addresses:
• Class B: 16 bits for
Network ID
• Total networks 214
• Network id starts with
‘10’ binary
• First byte: 128 to 191
◦

39
IP Addressing
• Classification of IPv4
addresses:
• Class C: 24 bits for
Network ID
• Total networks 221
• Network id starts with
‘110’ binary
• First byte: 192 to 223
◦

40
IP Addressing
• Classification of IPv4
addresses:
• Class D: Used for
multicasting
• No prefix or Network ID
• First byte: 224 to 239

41
IP Addressing
• Classification of IPv4
addresses:
• Class E: Reserved for
future use
• First byte: 240 to 255

42
IP Addressing
• Address Masking:
• Classful addressing lead
to depletion of IP
addresses and/or
unused addresses.
• Solution:
• Classless addresses with
variable sized prefix
according to the needs
of organizations

43
IP Addressing
• Solution:
• A notation representing
the length of prefix is
added at the end of a
classless address with a
slash ‘/’ to indicate the
addresses in a classless
address block.

44
IP Addressing

• Module 18

45
IP Addressing
• Dynamic Host
Configuration Protocol
(DHCP) is used to
automatically assign IP
addresses from an
acquired block of IP
addresses.

46
IP Addressing

• Organizations use private

IP addressing for the LAN
devices and can use
Network Address
Translation (NAT)
mechanism for having a
single or a few registered
global IP address/es for
internet communication.

47
IP Addressing
• NAT enabled router
replaces the local address
of sending device with
the registered global IP
address before sending
the packets on internet.

48
IP Addressing
IP Addressing:
◦ The mapping of incoming
internet packets is done
through NAT table which
contains the source
device local address port
number of the program
along with
corresponding IP address
of internet device.

49
IP Addressing
IP Addressing:
◦ The internal network is
supposed to initiate the
internet communication
in NAT mechanism for
mapping to take place.

50
IP Addressing
• Anytime a host or a router
needs to find the link-layer
address of another host or
router in its network, it
broadcasts an Address
Resolution Protocol (ARP)
request packet with the
destination IP address and
its own IP and link level
address.
• The destination device
replies to the sender device
with its link level address.

51
IP Addressing
• As compared to IPv4, the
next generation IP
protocol is IPv6. Some
important changes are:
• No more NAT (Network
Address Translation)
• No more private
address collisions

52
IP Addressing
• As compared to IPv4, the
next generation IP
protocol is IPv6. Some
important changes are:
• Built-in authentication
and privacy support
• Easier administration
(no more DHCP
required)
• Simplified routing

53
IP Addressing
IP v6 Addressing:
• There are three
categories of IPv6
addresses:
• Unicast address: For a
single connection
• Multicast address: For a
set of interfaces, one
message transmitted to
all.

54
 IP Addressing
IP v6 Addressing:
• Anycast address: For a
group of interfaces, one
message transmitted to a
single interface

end

55
Cloud Computing
Module 19

56
Ethernet
• It is a popular LAN
technology for data-link
and physical layers.
• Institute for Electrical and
Electronic Engineers
(IEEE) developed an
Ethernet standard known
as IEEE Standard 802.3

57
Ethernet
• TCP/IP does not specify
any protocols for data-link
and physical layers. It
accepts all the protocols
working at these layers.

58
Ethernet
• Ethernet was developed
in 1970s and since then it
has gone through four
generations. This
evolution is in fact the
reason of vast
implementation of
Ethernet in the world.

59
Ethernet
• Data rate 10 Mbps
• Connectionless
• No flow control
• No error control
• No retransmission and
acknowledgement
• Hence unreliable like IP
and UDP

60
Ethernet
• Uses link-layer addresses
(the 48 bit MAC address)
• CRC is present but
corrupted frames are
simply discarded by
receiver
• Each frame is of 64-1518
bytes of length including
46-1500 bytes of data
• CSMA/CD is used

61
Ethernet
• Unicast address:
Significant bit of first byte
is 0
• Multicast address:
Significant bit of first byte
is 1

62
Ethernet
• Broadcast address: All 48
bits are 1s.
• Note: All devices on
Ethernet receive all the
messages but keep only
those that are addressed
according to above.

63
Ethernet
• Standard Ethernet types:
• Bridged
• Switched
• Full duplex switches

64
Ethernet
• Fast Ethernet:
• Next generation of
standard Ethernet
• Raised speed to 100
Mbps
• Downward compatible
with standard Ethernet
(speed is reduced for
compatibility)

65
Ethernet
• Fast Ethernet:
• Same 48 bit addressing
• Frame format is same as
of standard Ethernet
• Uses star topology for
connecting three or
more devices using
switch or hub

66
Ethernet
• Gigabit Ethernet:
• 1Gbps speed
• Compatible with
standard and fast
Ethernet
• Star topology using hub
or switch
• Up to 5 kilometers range

67
 Ethernet
• 10 Gigabit Ethernet:
• 10 Gbps speed
• Compatible with
standard and fast
Ethernet
• Increases the range to
tens of kilometers
• Possibility to
interconnect LANs

end

68
 Cloud Computing
Module 20

end

69
Wired LAN vs. Wireless LAN

Wired LAN Vs. Wireless LAN

70
 Wired LAN vs. Wireless LAN
WIRED LAN WIRELESS LAN
◦ Medium: Wires ◦ Medium: Air
◦ Broadcasting and ◦ All devices are
multicasting possible broadcasting
when required ◦ No physical connection
◦ Physical connection to to network
network ◦ No link layer switch exists
◦ Hosts are connected ◦ Connected to other
through link layer switch networks through access
◦ Connection to other point (a device that
networks through router connects a wireless and
wired network)

71
IEEE 802.11
• It is a wireless LAN
standard by IEEE that
covers physical and data-
link layers
• Synonyms: WiFi, Wireless
LAN
• Basic architecture
consists of an access
point (AP) and capable
devices connected to AP

72
IEEE 802.11
• In the absence of AP, the
wireless devices connect
to form adhoc network
• Multiple overlapping APs
are used to cover a larger
area
• A device is connected to
only one of the nearest
APs

73
IEEE 802.11
◦ CSMA/CA is used. The
sender sends a Request
To Send (RTS) packet, the
receiver sends Clear To
Send (CTS) packet, the
sender sends data after
receiving CTS, the
receiver sends
acknowledgement, the
other senders can send
now.
◦ If no CTS is received, the
sender marks it as a
collision

74
 IEEE 802.11
◦ 802.11 a, b, g, n
◦ 802.11a: 50 feet, 22 Mbps
◦ 802.11b: 100 feet, 11 Mbps
◦ 802.11g: 100 feet, 54 Mbps
◦ 802.11n: 50 feet, 700
Mbps (to be
implemented)

end

75
Cloud Computing
Module 21

76
Bluetooth
• It is a wireless LAN
technology that provides
short distance
connectivity to devices
which have different
functionalities for
example, mobile phones,
headsets, notebooks,
desktops, computer
peripheral devices,
cameras and even the
home appliances.

77
Bluetooth
• Multiple devices can be
connected through
Bluetooth to form a
piconet
• Bluetooth supports:
• Voice and data
transmission
• Adhoc networking for
up to 10 meters

78
Bluetooth
• Multiple devices can be
connected through
Bluetooth to form a
piconet
• IEEE standard 802.15
covers the Personal Area
Network (PAN) using
Bluetooth for an area
covering a room

79
Bluetooth

• Versions:
• 1.x: up to 1Mbps,
obsolete
• 2.x: up to 3 Mbps,
improved pairing
capability between
devices from different
manufacturers

80
Bluetooth

• Versions:
• 3.x: up to 24 Mbps using
WiFi 802.11
• 4.x: Up to 24 Mbps,
works seamlessly with
4G, works with data
collection from sensors
an internet of things
(IoT)

81
Cloud Computing
Module 22
WiMAX
• WiMax stands for
Worldwide
Interoperability for
Microwave Access
• Provides wireless access
to Internet for:
• Homes and offices
when the wired access
is either not available or
is expensive (fixed
WiMAX)
• Mobile phones (mobile
WiMAX)

83
WiMAX
• Fixed WiMAX requires the
installation of antennas at
the premises of the
subscriber to receive and
send the data from the
base station of Internet
provider.

84
WiMAX
• Mobile WiMAX users
move from one place to
another while connected
to the base station of
Internet provider .

85
WiMAX
• WiMAX is the result of
IEEE 802.16 project. It is a
standard for wireless
WAN (or MAN). The
subscriber station may be
tens of kilometers away
from the base station of
the provider.
• Remember that 802.11 is
the standard for wireless
LAN.

86
WiMAX
• Uses 48-bit MAC address
of subscriber station and
base station at Data-link
layer
• Connection oriented
protocol. Each connection
has a unique id and hence
there is no address field in
the frame of WiMAX
• Full duplex
communication

87
Cloud Computing
Module 23

88
Evolution of Cellular Networks
• Cellular network or
telephony is a radio-based
technology
• Radio waves are
electromagnetic waves
propagated by antennas
• Note: Antenna is a
transducer device which
converts the altering
current into radio waves
and vice versa

89
Evolution of Cellular Networks
• 7 billion mobile
connections
• 25 billion interconnected
devices count predicted in
2020
• Over 100 billion
downloads completed in
2013, 270 billion expected
in 2017

90
Evolution of Cellular Networks
• The base stations receive
from and transmit to
cellular phones.
• Cellular Networks have
evolved from first
generation (1G) to fifth
generation (5G)
• Let us briefly look at these
generations…

91
Evolution of Cellular Networks

• 1G
• Invented around 1980.
• First implementation in
Tokyo (Japan)
• Based upon analog
technology
• Expanded to cover all
the population of Japan
in few years

92
Evolution of Cellular Networks

• 1G
• Not secure
• Anyone with an all-band
radio can listen to calls
and get the phone
number of the
subscriber
• Analog mobiles were
larger in size and heavy
in weight

93
Evolution of Cellular Networks

• 2G
• Invented in 1991,
implemented first time in
Finland
• Technologies: Global
System for Mobile (GSM)
Communication, General
Packet Radio Service
(GPRS), Code Division
Multiple Access (CDMA)
[digital signal] and
Enhanced Data Rates for
GSM Evolution (EDGE)

94
Evolution of Cellular Networks

• 2G
• Short Messaging Service
(SMS), Multi-Media
Messaging Service (MMS
• Typical data rate: 100
Kbps
• Email, Web browsing,
Camera phones

95
Evolution of Cellular Networks

• 2G
• Signal strength decay
problem, performance
degrades with the rise in
number of users in a cell
(area maintained by a
base station)

96
Evolution of Cellular Networks

• 3G
• From 2000 to 2010
• Technologies: CDMA,
WLAN, Bluetooth,
Universal Mobile
telecommunication
Systems (UMTS), High
Speed Downlink Packet
Access (HSDPA)

97
Evolution of Cellular Networks

• 3G
• Features: Global
Roaming Clarity in voice
calls, Fast
Communication,
Internet, Mobile T.V,
Video Conferencing,
Video Calls, Multi Media
Messaging Service
(MMS), 3D gaming
and Multiplayer-
Gaming, smart phones

98
Evolution of Cellular Networks

• 3G
• Typical data rate: Up to a
few Mbps
• Expensive mobile
phones, battery life issue

99
Evolution of Cellular Networks

• 4G
• Since 2010
• Technologies: Long
Term Evolution (LTE)
Standard based on the
GSM/EDGE and
UMTS/HSPA, Multiple In
Multiple Output (MIMO)
smart antenna
technology, Orthogonal
Frequency Digital
Multiplexing (OFDM),
WiMAX

100
Evolution of Cellular Networks

• 4G
• Typical data rate: Up to
a few tens of Mbps
• MAGIC: Mobile
multimedia–Anytime
anywhere–Global
mobile support–
Integrated wireless
solutions–Customized
personal service

101
Evolution of Cellular Networks

• 4G
• Maintaining data rate is
an issue, not fully
implemented in all the
world, battery
consumption is a bigger
problem than 3G

102
Evolution of Cellular Networks

• 5G
• To be implemented
• Technologies: New
releases of LTE
• Faster data rate than 4G
(> 1Gbps), higher data
rate at cell edges
• Research is still in
progress

103
Cloud Computing
Module 24
Connecting Devices
• Connecting Devices:
These are the devices
used to connect:
• Hosts to form LANs
• LANSs to implements
WANs and Internet

105
Connecting Devices
• The class of each device
depends upon the layer/s
on which it operates. That
is:
• Hub: Physical layer
• Link-layer Switch:
Physical layer, Data-link
layer
• Router: Network layer,
Data-link layer, Physical
layer

106
Connecting Devices
• Hub: It is a multiport
repeater device used in
star topology.
• A repeater device
regenerates the signal
before it become too
weak or corrupted.

107
Connecting Devices
• Hub: The hub repeats the
signal received from any
port ‘A’ to all the other
ports except the port ‘A’
(broadcasting)
• This is because the hub
is a physical layer
device. It does not has
its own MAC address
and can not access the
data-link layer address
(MAC address) of the
data frames.

108
Connecting Devices
• Switch: It is a multiport
bridge device.
• A bridge joins two
logical segments of the
same network and
intelligently forwards
the packets from one
segment to other on the
basis of destination
MAC address and
forwarding table.

109
Connecting Devices
• Switch: It is a two layer
device. It performs
functionality on data-link
layer as well as it
regenerates the signals it
receives.

110
Connecting Devices
• Switch: A data-link layer
switch works by
maintaining a switching
table and forwarding the
packets received from a
port ‘A’ only to the
destination port ‘B’.

111
Connecting Devices
• Switching table consists
of MAC address of the
hosts arranged according
to the port numbers to
which they are attached.
It is consulted before
forwarding a packet.

112
Connecting Devices
• Switch: Switches
perform learning to fill
the switching table by
reading the MAC address
of sending device for each
port.
• Broadcast packets are
forwarded to all ports

113
Connecting Devices
• Switch: In a situation
when two LANs are
connected through more
than one switch then the
looping problem can
arise.

114
Connecting Devices
• Looping: The frame/s
sent by one host ‘X’ in
LAN1 to a host ‘Y’ in LAN2
will go through multiple
switches and get
duplicated when reaching
the destination host.
• Reason: The switches
generically forward the
frame received for an
unregistered MAC to all
the ports (except the
sender’s port).

115
Connecting Devices
• Looping: The frame/s
sent by one host ‘X’ in
LAN1 to a host ‘Y’ in LAN2
will go through multiple
switches and get
duplicated when reaching
the destination host.
• Solution: The switches
should be connected in a
virtual topology to avoid
any loop or cycle; the
Spanning tree
arrangement.

116
Connecting Devices
• Advantages of
Switch:
• Collision elimination
• Connecting
heterogeneous devices
(in terms of data rate
capacity)

117
Connecting Devices
• Router: It is a three layer
device:
• Physical (regenerating
the signals)
• Data-link layer(checking
the MAC addresses of
source and destination)

118
Connecting Devices
• Router: It is a three layer
device:
• Network layer (checks
the IP addresses of
source and destination,
connects multiple
networks to form
bigger networks)

119
Connecting Devices
• Router: Has multiple
interfaces. Each interface
has a MAC address and IP
address.
• A router:
• Only processes those
packets which are
addressed to the
interface at which they
arrive.

120
Connecting Devices
• Router: Has multiple
interfaces. Each interface
has a MAC address and IP
address.
• A router:
• Changes the source and
destination MAC
address when it
forwards the packets.

121
Connecting Devices
• Virtual LAN (VLAN): A
logical (not physical)
segment of a physical
LAN.
• VLANs are defined by
software. Each VLAN is a
work group in an
organization, has a VLAN
ID and receives the
broadcast messages
addressed to its own ID.

122
 Connecting Devices
• A VLAN may span over
multiple switches in a
LAN.
• No need to update the
physical topology to
relocate a person from
one VLAN to other, just
the software
configuration is to be
END
updated.

123
Cloud Computing
Module 25

1
Routing
• In a physical network,
multiple LANs and
WANs are joined
together by the routers.
• Hence there can be
more than one route
between two hosts.
• Routing is a service of
Network layer to find
the best route.

2
Routing
• Routing is performed by
applying routing
protocols and using the
decision tables called
routing tables in each
router.
• Forwarding is the action
performed by a router
on the basis of routing
protocol and routing
table according to the
destination address of
each packet received at
any interface.

3
Routing
• At network layer, each
message from higher
layer is broken down
into packets.
• A router performs
packet switching.

4
Routing
• Types of routing:
• Unicast routing: A
router forwards the
packet to only one of
the attached
networks.
• Multicast routing: A
packet is forwarded to
multiple attached
networks.

5
Routing
• Routing a packet from a
source host to
destination host can
also be defined as
routing a packet from a
source router (the
default router of the
source host) to a
destination router (the
router connected to the
destination network)
through the
intermediate routers
using routing
algorithms.
6
Routing
• Types of routing:
• Connectionless routing:
All packets of the
same message are
treated independently
and may or may not
follow the same route.
• Connection oriented
routing: All the packets
of same message are
labeled and routed
through a virtual
circuit or a fixed route.

7
 Routing
• An internet can be
considered as a graph
with each network as an
edge and each router as
a node.
• In a weighted graph,
each edge has a weight
or cost.
• Least cost routing can
be performed. Example
end algorithms: Distance-
Vector routing, Link-
State routing

8
Cloud Computing
Module 26

9
Broadband Networks & Internet
• All clouds are inherently
dependent upon
internetworking or
Internet for ubiquitously
remote provisioning of
IT resources.
• The cloud providers and
consumers connect to
Internet through ISPs.

10
Broadband Networks & Internet
• The largest backbone
networks of the
Internet are
strategically
interconnected by core
routers.
• The core-routers
connect the
international networks.

11
Broadband Networks & Internet
• The Internet has
become a dynamic and
complex aggregate of
ISPs.
• There is a hierarchical
topology for worldwide
interconnectivity
composed of tiers.

12
Broadband Networks & Internet
• There are three tiers of
worldwide connectivity:
• Tier 1 consists of large-
scale international
connectivity providers.
• Tier 2 consists of large
regional ISPs
connected to tier 1.
• Tier 3 consists of local
ISP providers
connected to tier 2.

13
Broadband Networks& Internet
• The cloud providers
and users connect
directly to tier 3
providers.

end

14
Cloud Computing
Module 27

15
Internet Architecture & Cloud deployment
• Internet supports the
remote provisioning of
IT resources.
• Cloud relies heavily
upon Internet.

16
Internet Architecture & Cloud deployment
• The connectivity of
end-users of cloud
depends upon how
the centralized
resources of cloud are
deployed.
• The cloud resources
deployment can be
either on-premises or
Internet based.

17
Internet Architecture & Cloud deployment
• In cloud deployment
using the on-premises,
the provider sets up a
fully controlled
corporate network
and a corporate
Internet connection
for the deployment of
IT solutions and
applications.

18
Internet Architecture & Cloud deployment
• In the on-premises
deployment, the
internal users access
the cloud through
corporate network.
The remote users
connect through
internet by using
virtual private network
(VPN).

19
Internet Architecture & Cloud deployment
• A VPN creates a secure
connection between a
remote device and the
corporate servers over
the internet as if the
device is inside the
LAN.

20
Internet Architecture & Cloud deployment
• For the internet based
deployment, the cloud
provider has an
Internet connection
and al the internal and
external users access
the cloud resources
through cloud
provider’s internet
connection.
end • In this deployment,
there is an extra
charge for internet
connectivity.

21
Cloud Computing
Module 28

22
Scalable Computing over Internet
• Scalable computing may
refer to the dynamic
resizing of the available
computing resources
(processing, memory,
bandwidth, storage
etc.) with demand.

23
Scalable Computing over Internet
• The growth of users and
user demands for
scalable computing over
internet has been
accompanied with
matching growth in
network, computing
and resource
management
technologies.

24
Scalable Computing over Internet
• The computing
platforms have evolved
as follows
• Mainframes (1950-70)
• Minicomputers (1960-
1980)
• Personal computers
(1970-1990)
• Portable computers
(1980-2000)

25
Scalable Computing over Internet
• Since 1990, the High
Performance
Computing (HPC) and
High Throughput
Computing (HTC) have
been relying upon
clusters, grids and the
Internet clouds.

26
Scalable Computing over Internet
• The speed for HPC
systems
(supercomputers) has
increased from Gflops
in early 1990s to now
Pflops in 2010.

27
Scalable Computing over Internet
• The network
bandwidth has been
doubling each year in
the recent past
(Gilder’s law).
• Processor speed has
been doubling every 18
months (Moore’s law).
• Means that there has
been a steady growth
in these technologies.

28
Scalable Computing over Internet
• Fine grain (instruction
level) parallelism and
coarse grain (job level)
parallelism are
available.

29
Scalable Computing over Internet
• Ubiquitous computing
is what refers to
computing at any
place and time using
pervasive devices and
wired or wireless
communications.
• Utility computing
works upon a business
model in which the
customers pay for
computational
resources from a
provider.

30
Scalable Computing over Internet
• Cloud computing
provides ubiquitous
utility computing.

end

31
Cloud Computing
Module 29

32
Technologies for Network based Systems
• The processor speed
and network
bandwidth have
shown a remarkable
growth in last few
decades.
• The processor clock
rate has risen from 10
MHz in 1970s to over
4GHz in 2010s.

33
Technologies for Network based Systems
• The network band has
increased from 10
Mbps to over 100,000
Mbps

34
Technologies for Network based Systems
• The excessive heat
generation from single
processor core with
high frequency has
limited the maximum
speed unless the chip
technology matured.

35
Technologies for Network based Systems
• This has lead to the
multi-core architecture
of CPUs with dual,
quad, six or more
cores.
• The graphical
processing unit (GPU)
development has
adopted a many-core
architecture with
hundreds to
thousands of cores.

36
Technologies for Network based Systems
• Modern architecture
of CPUs and GPUs
have enhanced the
instruction level
parallelism (ILP) and
the volume of millions
of instructions per
second (MIPS).
• Sun’s Niagara CPU can
provide 64 count for
ILP.
• Intel’s Core i7 990x can
provide 159,000 MIPS
execution rate
37
Technologies for Network based Systems
• The CPUs and GPUs
are multithreaded,
which means that each
core can execute
multiple processes or
threads concurrently.
• A GPU unit has far
more (but slower)
cores than a multi-core
CPU.

38
Technologies for Network based Systems
• The DRAM memory
chip capacity has
increased from 16 KB
in 1976 to 64 GB in
2011.
• The hard disk capacity
has increased from
260 MB in 1981 to 3TB
a few years ago.
• The flash memory and
solid state drives are
rapidly evolving.

39
Technologies for Network based Systems
• Disk arrays are being
utilized to enhance the
storage.
• Servers can be
connected to network
storage such as disk
arrays through storage
area network (SAN)
• A disk array can be
connected to client
hosts through network
attached storage
(NAS)

40
Technologies for Network based Systems
• The high bandwidth
networks in WAN
scope can connect the
host computers to
network storage.
• A single host can be
shared among multiple
instances of operating
systems through
virtualization
end technology. More on
this latter.

41
Cloud Computing
Module 30

42
Web 2.0
• It is the second
generation of world
wide web.
• Lets people
collaborate and share
comments, media and
information online.

43
Web 2.0
• The web pages
progressed from static
to dynamic and
interactive.
• Through
Asynchoronous
Javascript and XML or
Ajax, the web
applications can send
and receive data from
a web server without
interfering with the
display and behavior
of the existing page.

44
Web 2.0
• Social networking and
community oriented
sites have emerged
• For example
myspace.com,
facebook.com,
twitter.com etc.

45
Web 2.0
• Users can contribute in
web based blogs,
wikis, online training,
online education etc.
• Really Simple
Syndication (RSS)
feeds continuously
keep the subscribers
informed about news,
follow up updates and
products.

46
 Web 2.0
• Users can do online
chatting and share
files through
messenger tools such
as Yahoo messenger,
Skype, WhatsApp etc.

end

47
Cloud Computing
Module 31

48
Virtualization
• Virtualization: It is a
technology used to
enhance the utilization of
computing resources.
• A single hardware
machine is multiplexed
among multiple virtual
machines (VMs).

49
Virtualization
• A software based virtual
machine
monitor/manager (VMM)
or hypervisor is a
program that manages
the hardware resources
for the VMs and also
keeps each VM from
disrupting other VMs.

50
Virtualization Levels
• Virtualization
implementation levels:
1. Instruction Set
Architecture (ISA) level:
Executing legacy code
over new machines
using ISA emulator tool
such as an interpreter
which translate one
instruction of source
code into corresponding
instruction of the target
machine.

51
Virtualization Levels
• Virtualization
implementation levels:
2. Hardware Abstraction
level:
The hardware
components (CPU,
RAM, Disk, NIC) of a
physical system are
virtualized and shared
among virtual machines
using Virtual Machine
Monitor (VMM) tool or
hypervisor which
performs as abstraction
layer.

52
Virtualization Levels
• Virtualization
implementation levels:
3. Operating System
Level:
The OS running over a
server accommodates
multiple containers or
VMs. The host operating
system acts as the
abstraction layer
between hardware and
the containers.

53
Virtualization Levels
• Virtualization
implementation levels:
4. Library support level:
The API calls for
hardware acceleration
such as vCUDA stubs for
graphic processing units
(GPUs) are available at
VM level.

54
Virtualization Levels
• Virtualization
implementation levels:
5. Application level:
An application acts as a
VM through wrapping of
application in an
abstraction layer which
isolates it from OS and
other applications.
Another type is using
virtualization layer as
programming
environment e.g; Java
Virtual Machine (JVM).

55
Virtualization
• Hypervisor transforms
the physical hardware
into virtual hardware.

56
Virtualization
• Virtualization
Architecture

VM 1 VM 2 VM 3

Hypervisor

Hardware

57
Virtualization
• Hypervisor Architecture:
• Monolithic

VM 1 VM 2 VM 3

Hypervisor
Drivers

Hardware

58
Virtualization
• Hypervisor Architecture:
• Microkernel

VM 1 VM 2 VM 3
Drivers Drivers Drivers

Hypervisor

Hardware

59
Virtualization
• Full Virtualization

VM 1 VM 2 VM 3
(unmod (unmod (unmod
ified ified ified
guest guest guest
OS) OS) OS)

Hypervisor

Hardware

60
Virtualization
• Para Virtualization

VM 1 VM 2 VM 3
(modifie (modifie (modifie
d guest d guest d guest
OS) OS) OS)

Hypervisor

Hardware

61
Virtualization
• Native Virtualization
Architecture

VM 1 VM 2 VM 3

Hypervisor

Hardware

62
Virtualization
• Hosted Virtualization
Architecture

VM 2 VM 3

App Hypervisor

Host Operating System

Hardware

63
Cloud Computing
Module 32

64
VM Architectures
• We know that the
virtualization layer
transforms the physical
hardware into virtual
hardware. There are three
classes of VM
architectures.

65
VM Architectures
1. Hypervisor Architecture
2. Full-virtualization
Architecture
3. Para-virtualization
Architecture

66
VM Architectures
1. Hypervisor Architecture
• It is the hardware level
virtualization. Also called
the bare-metal
virtualization
• The hypervisor sits
between the hardware
and the VMs and
manages the VMs.
• Example: Xen, VMware

67
VM Architectures
2. Full Virtualization: The
guest operating system
(OS) or the VM’s OS does
not know that it is installed
on a VM.
The Virtualization layer
manages the hardware
acceleration. For example
VMware
The virtualization layer can
be installed on hardware or
on host’s OS.
Some of the instructions of
a gust VM are directly run
on hardware to enhance
the performance.
68
VM Architectures
3. Para-virtualization
Architecture: The guest
OS is modified to comply
with virtualization layer.
All calls for hardware
acceleration are handled
by virtualization layer.
For example: KVM

end

69
Ple

70
Cloud Computing
Module 24

71
Basics: The Internet
Network
◦ Internet Architecture:
◦ Backbones: Large
networks owned by
communication
companies such as PTCL,
AT&T etc.
◦ Backbone networks are
joined at peering points.
Basics: The Internet
Network
◦ Internet Architecture
◦ Provider Networks: Use
the service of backbone
for a fee. Connected to
backbone through
peering points.
◦ Sometimes connected to
other provider networks
as well.
Basics: The Internet
Network
◦ Internet Architecture
◦ The Backbones and
provider networks are
also called Internet
Service Providers (ISPs).
Basics: The Internet
Network
◦ Internet Architecture
◦ Customer Networks:
Are the networks at the
edge of Internet.
◦ They pay fee to
Providers for using the
Internet services.
◦ They are the actual
users of Internet
services.
Cloud Computing
Module 33

1
Virtualization Figures
• Hypervisor transforms
the physical hardware
into virtual hardware.

2
Virtualization Figures
• Virtualization
Architecture

VM 1 VM 2 VM 3

Hypervisor

Hardware

3
Virtualization Figures
• Hypervisor Architecture:
• Monolithic

VM 1 VM 2 VM 3

Hypervisor
Drivers

Hardware

4
Virtualization Figures
• Hypervisor Architecture:
• Microkernel

VM 1 VM 2 VM 3
Drivers Drivers Drivers

Hypervisor

Hardware

5
Virtualization Figures
• Full Virtualization

VM 1 VM 2 VM 3
(unmod (unmod (unmod
ified ified ified
guest guest guest
OS) OS) OS)

Hypervisor

Hardware

6
Virtualization Figures
• Para Virtualization

VM 1 VM 2 VM 3
(modifie (modifie (modifie
d guest d guest d guest
OS) OS) OS)

Hypervisor

Hardware

7
Virtualization Figures
• Native Virtualization
Architecture

VM 1 VM 2 VM 3

Hypervisor

Hardware

8
Virtualization Figures
• Hosted Virtualization
Architecture

VM 2 VM 3

App Hypervisor

Host Operating System

Hardware
end

9
Cloud Computing
Module 34

10
Virtualization of CPU, Memory and I/O Devices

• In order to support
virtualization,
processors such as x86
architecture use a
special mode and
instructions known as
hardware-assisted
virtualization.

11
Virtualization of CPU, Memory and I/O Devices

• In this way, the

hypervisor is able to
trap the sensitive
instructions of the
guest OS and its
applications.

12
Virtualization of CPU, Memory and I/O Devices

• The modern processors

allow multiple
processes to run
simultaneously. Any
process can execute a
critical instruction to
crash the whole system.
• Therefore the critical
instructions are
executed in privileged or
supervisor mode of the
processor. The OS
controls this mode on
behalf of the processes
being executed.
13
Virtualization of CPU, Memory and I/O Devices

• The second type of

instructions are non-
privileged or non-critical
instructions which are
run in user-mode of the
processor.

14
Virtualization of CPU, Memory and I/O Devices

• CPU Virtualization: A
CPU is virtualizable if it
is able to run the
privileged and un-
privileged instructions
of a VM in user mode
and the hypervisor
running in supervisor
mode.

15
Virtualization of CPU, Memory and I/O Devices

• Memory Virtualization:
Traditionally, the OS
performs the mapping
of virtual memory to
machine memory by
using page tables.
• The modern x86 CPUs
include the memory
management unit
(MMU) and translation
lookaside buffer (TLB) to
optimize virtual memory
performance.

16
Virtualization of CPU, Memory and I/O Devices

• However in
virtualization
environment, the
memory virtualization
involves the sharing and
dynamic allocation of
physical memory of the
system to the physical
memory of the VMs.

17
Virtualization of CPU, Memory and I/O Devices

• The guest OS performs

the virtual to physical
memory mapping of the
VM, while the
hypervisor performs
mapping of physical
memory to machine
memory.

18
Virtualization of CPU, Memory and I/O Devices

• I/O Virtualization:
• It is done in either of
the three ways:
1. Full device
emulation: The
device is emulated in
software located in
hypervisor. The
hypervisor interacts
with the real device.
The VM interacts
with the virtual
device.

19
Virtualization of CPU, Memory and I/O Devices

• Para-virtualization
based I/O: The guest OS
interacts with the
device through its
frontend driver. The
frontend driver
interacts with a backend
driver of the device. The
backend driver interacts
with the device.

20
Virtualization of CPU, Memory and I/O Devices

• Direct I/O virtualization:

This type of I/O
virtualization allows the
VMs to directly access
the device.

end

21
Cloud Computing
Module 35

22
Virtual Clusters
• A virtual cluster consists
of several VMs hosted
on a physical cluster.
• The VMs are
interconnected through
a virtual network across
multiple physical
networks.
• The nodes can be
physical or virtual
machines and can grow
or shrink dynamically.

23
Virtual Clusters
• The failure of a host can
make the hosted VMs
unavailable but the
virtual cluster does not
collapse.
• The failure of a VM does
not fails the host.
• A physical cluster may
host multiple virtual
clusters.
• A virtual cluster may
span over multiple
physical cluster.

24
Virtual Clusters
• In order to deploy a
virtual cluster, several
VMs with installed OS
and application
software are required.
• The deployment time is
to be as quick as
possible.

25
Virtual Clusters
• Templates can be used
to deploy the VMs from.
• A template is a disk
image with preinstalled
OS with or without
certain applications.
• A suitable template can
be copied as disk image
of a VM. This saves time
of installing and
configuring.

26
Virtual Clusters
• When the VM is ready
and up, it is deployed to
a suitable host.
• The VM then joins a
virtual cluster.
• All of the above can be
done manually as well
as full or partially
automated.

27
Virtual Clusters

• Reasons of
virtualization:
1. Sharing of
resources
2. Isolation of users
of shared resource
3. Aggregation of
smaller resources
into a single big
virtual resource
(e.g., Storage)

28
 Virtual Clusters

• Reasons of
virtualization:
4. Dynamic
relocation/provisio
ning of virtual
resources is easier
than physical
resources
5. Easier
end management of
virtual
resources/devices/
machines.
29
Cloud Computing
Module 36

30
VM Migration
• VMs can be migrated
from one host to
another for:
1. Server Load
balancing
2. Server consolidation
3. Remedy for failover
hosts and VMs
4. Remedy for resource
shortage for a VM

31
VM Migration
• A VM can be in any of
the following states:
1. Powered-off
2. Suspended
3. Paused
4. Powered-on

32
VM Migration
• The following options
are available for VM
migration:
1. Cold migration: The
VM has to be
powered-off before
migration.
2. Warm migration:
Suspended VM
migration.
3. Live migration: For
powered-on VM with
zero downtime and
full availability.
33
VM Migration
• A VM is made of two
basic components:
• VM state: The
processor and RAM
contents
• Virtual hard disk:
Residing on network
storage or on host’s
hard disk
• Live migration of VM
means zero downtime
of OS, connectivity and
applications running on
the VM.
34
VM Migration
• For live migration, the
VM state is to be copied
from source to
destination host. The
virtual disk can also be
migrated through live
storage migration
feature of the
hypervisor.
• Modern day hypervisors
provide unbroken
network connectivity of
the VM during live
migration.

35
VM Migration
• During the live
migration, the state and
storage of the VM keeps
on working on source
host to avoid down-
time.
• For live migration of a
VM with the virtual hard
disk on network
accessed shared
storage, the virtual hard
disk need not to be
copied if the destination
host can access that
network based storage.
36
 VM Migration
• Migrating the virtual
hard disk is time
consuming as well as
network bandwidth
consuming due to multi-
Gigabyte migration.
• A better solution is to
use the network
storage.

end

37
Cloud Computing
Module 37

38
Virtualization for Data Center Automation
• A data center is a facility
with networked
computers and is used
by businesses and other
organizations to
process, store and share
large amounts of data.
• Companies like Google,
Yahoo, Amazon,
Microsoft, IBM, HP,
Apple etc. have invested
billions of dollars for
constrcting the data
centers.

39
Virtualization for Data Center Automation
• Data center automation
refers to the dynamic
provisioning of
hardware and software
resources to millions of
users simultaneously.
• Data centers can host
Clouds.
• Data center automation
is triggered by the
growth of virtualization
products.

40
Virtualization for Data Center Automation
• The data center owner
has three major
considerations:
1. Assuring
Performance and
QoS
2. Increase resource
utilization
3. Saving costs

41
Virtualization for Data Center Automation
• Enhanced resource
allocation (to jobs
and/or VMs) may be
performed in data
centers to assure
performance and QoS.

42
Virtualization for Data Center Automation
• The over allocation of
computing resources
may result in decrease
in average utilization of
these resources.
• This also leads to
increased costs due to
power consumption.

43
Virtualization for Data Center Automation
• Example: A VM hosted
on a server with 1.5 GHz
*4 cores and 16 GB of
RAM is allocated 1.5GHz
* 2 vCPUS, 4 GB vRAM
(half of the processing
and 1/4th RAM).

44
Virtualization for Data Center Automation
Suppose if there are two
such VMs. But the overall
average workload of the
hosted VMs keeps the
physical utilization to less
than 50%. This is a
resource wastage as 50%
of the resources remain
idle.

45
Virtualization for Data Center Automation
• Server consolidation is a
technique by which
more VMs are
aggregated on a single
server (by migrating
jobs/VMs to it) while
assuring performance
and QoS.
• This increases the
resource utilization
across data center.
• More servers are
available to take more
workload.
46
Virtualization for Data Center Automation
• More servers are
available to take more
workload. Otherwise,
the idle servers can be
shut down to save
power.
• Virtualization
technology also helps in
setting of virtual
storage (over VMs) to
offer virtual disks to
other VMs.

47
Virtualization for Data Center Automation
• Virtualization can
synchronize with cloud
management systems
to dynamically provision
cloud services and
billing systems.
• Hence, virtualization is
essential for Cloud
computing.

end

48
Cloud Computing
Module 38

49
Virtual Private Network (VPN)
• A VPN extends a private
network over public
network and enables
the users to
communicate as if their
devices are directly
connected to the
private network.
• A VPN creates a secured
and encrypted network
over a less secured
network such as the
Internet.

50
Virtual Private Network (VPN)
• Normally a VPN is
provided and managed
by a service provider.
• VPN allows the
corporate employees to
securely access the
applications hosted
over enterprise LAN.
• VPN is based upon IP
tunneling.

51
Virtual Private Network (VPN)
• IP tunneling or port
forwarding is the
transmission of private
network packets over a
public network
(Internet) as the
payload of public
network packets such
that the routing devices
do not come to know
about this.

52
Virtual Private Network (VPN)
• There are many
protocols for VPN
establishment and
encryption: IP Security
(IPSec), Secure Socket
Layer(SSL), Point-To-
Point Tunneling
Protocol (PPTP),
Multiprotocol Label
Switching (MPLS) etc.

53
Virtual Private Network (VPN)
• VPN although provide
secured connectivity to
extend a private
network but the
implementation may
have performance
issues.
• VPN is implementable
over Layer 1-3.

54
Virtual Private Network (VPN)
• Types of VPN:
1. Remote-access VPN:
A VPN client on
user’s device
connected to VPN
gateway of the
enterprise.
2. Site-to-site VPN:
Establishes a VPN
between two
networks over the
Internet by using
VPN gateway.

55
Virtual Private Network (VPN)
• VPN technology
provides access to cloud
resources. The VPN
gateway exists in the
cloud with a secure link
provided by the cloud
provider.

end

56
Cloud Computing
Module 39

57
Network Virtualization

58
 Network Virtualization

• Multiple virtual Network Interface Cards (vNIC)

are linked to physical NIC or pNIC through a
virtual Switch (vSwitch) inside a hypervisor.

VM 1 VM 2 VM 3

vNIC vNIC vNIC

vSwitch

pNIC

59
Network Virtualization
• A virtual network
consists of virtual nodes
and virtual links.
• Network virtualization
establishes the
coexistence of multiple
virtual networks.

60
Network Virtualization
• A virtual network
consists of virtual nodes
and virtual links.
• Network virtualization
establishes the
coexistence of multiple
virtual networks.

61
Network Virtualization
• Network virtualization
proposes the
decoupling of
traditional ISP
functionalities such as
infrastructure setup and
management from the
creation and
management of virtual
networks.

62
Network Virtualization
• It is possible to use
physical infrastructures
of multiple providers to
dynamically compose
virtual network/s.

63
Network Virtualization
• Technologies:
1. Virtual Local Area
Network (VLAN):
Logically grouping
the hosts with
common interest
into a single
broadcasting
domain.

64
Network Virtualization
• Technologies:
2. Virtual Private
Networks (VPN): A
dedicated
communications
network of
enterprise/s and user/s
by using tunneling
over public networks
(Internet).
end

65
Cloud Computing
Module 33

1
Virtualization Figures
• Hypervisor transforms
the physical hardware
into virtual hardware.

2
Virtualization Figures
• Virtualization
Architecture

VM 1 VM 2 VM 3

Hypervisor

Hardware

3
Virtualization Figures
• Hypervisor Architecture:
• Monolithic

VM 1 VM 2 VM 3

Hypervisor
Drivers

Hardware

4
Virtualization Figures
• Hypervisor Architecture:
• Microkernel

VM 1 VM 2 VM 3
Drivers Drivers Drivers

Hypervisor

Hardware

5
Virtualization Figures
• Full Virtualization

VM 1 VM 2 VM 3
(unmod (unmod (unmod
ified ified ified
guest guest guest
OS) OS) OS)

Hypervisor

Hardware

6
Virtualization Figures
• Para Virtualization

VM 1 VM 2 VM 3
(modifie (modifie (modifie
d guest d guest d guest
OS) OS) OS)

Hypervisor

Hardware

7
Virtualization Figures
• Native Virtualization
Architecture

VM 1 VM 2 VM 3

Hypervisor

Hardware

8
Virtualization Figures
• Hosted Virtualization
Architecture

VM 2 VM 3

App Hypervisor

Host Operating System

Hardware
end

9
Cloud Computing
Module 34

10
Virtualization of CPU, Memory and I/O Devices

• In order to support
virtualization,
processors such as x86
architecture use a
special mode and
instructions known as
hardware-assisted
virtualization.

11
Virtualization of CPU, Memory and I/O Devices

• In this way, the

hypervisor is able to
trap the sensitive
instructions of the
guest OS and its
applications.

12
Virtualization of CPU, Memory and I/O Devices

• The modern processors

allow multiple
processes to run
simultaneously. Any
process can execute a
critical instruction to
crash the whole system.
• Therefore the critical
instructions are
executed in privileged or
supervisor mode of the
processor. The OS
controls this mode on
behalf of the processes
being executed.
13
Virtualization of CPU, Memory and I/O Devices

• The second type of

instructions are non-
privileged or non-critical
instructions which are
run in user-mode of the
processor.

14
Virtualization of CPU, Memory and I/O Devices

• CPU Virtualization: A
CPU is virtualizable if it
is able to run the
privileged and un-
privileged instructions
of a VM in user mode
and the hypervisor
running in supervisor
mode.

15
Virtualization of CPU, Memory and I/O Devices

• Memory Virtualization:
Traditionally, the OS
performs the mapping
of virtual memory to
machine memory by
using page tables.
• The modern x86 CPUs
include the memory
management unit
(MMU) and translation
lookaside buffer (TLB) to
optimize virtual memory
performance.

16
Virtualization of CPU, Memory and I/O Devices

• However in
virtualization
environment, the
memory virtualization
involves the sharing and
dynamic allocation of
physical memory of the
system to the physical
memory of the VMs.

17
Virtualization of CPU, Memory and I/O Devices

• The guest OS performs

the virtual to physical
memory mapping of the
VM, while the
hypervisor performs
mapping of physical
memory to machine
memory.

18
Virtualization of CPU, Memory and I/O Devices

• I/O Virtualization:
• It is done in either of
the three ways:
1. Full device
emulation: The
device is emulated in
software located in
hypervisor. The
hypervisor interacts
with the real device.
The VM interacts
with the virtual
device.

19
Virtualization of CPU, Memory and I/O Devices

• Para-virtualization
based I/O: The guest OS
interacts with the
device through its
frontend driver. The
frontend driver
interacts with a backend
driver of the device. The
backend driver interacts
with the device.

20
Virtualization of CPU, Memory and I/O Devices

• Direct I/O virtualization:

This type of I/O
virtualization allows the
VMs to directly access
the device.

end

21
Cloud Computing
Module 35

22
Virtual Clusters
• A virtual cluster consists
of several VMs hosted
on a physical cluster.
• The VMs are
interconnected through
a virtual network across
multiple physical
networks.
• The nodes can be
physical or virtual
machines and can grow
or shrink dynamically.

23
Virtual Clusters
• The failure of a host can
make the hosted VMs
unavailable but the
virtual cluster does not
collapse.
• The failure of a VM does
not fails the host.
• A physical cluster may
host multiple virtual
clusters.
• A virtual cluster may
span over multiple
physical cluster.

24
Virtual Clusters
• In order to deploy a
virtual cluster, several
VMs with installed OS
and application
software are required.
• The deployment time is
to be as quick as
possible.

25
Virtual Clusters
• Templates can be used
to deploy the VMs from.
• A template is a disk
image with preinstalled
OS with or without
certain applications.
• A suitable template can
be copied as disk image
of a VM. This saves time
of installing and
configuring.

26
Virtual Clusters
• When the VM is ready
and up, it is deployed to
a suitable host.
• The VM then joins a
virtual cluster.
• All of the above can be
done manually as well
as full or partially
automated.

27
Virtual Clusters

• Reasons of
virtualization:
1. Sharing of
resources
2. Isolation of users
of shared resource
3. Aggregation of
smaller resources
into a single big
virtual resource
(e.g., Storage)

28
 Virtual Clusters

• Reasons of
virtualization:
4. Dynamic
relocation/provisio
ning of virtual
resources is easier
than physical
resources
5. Easier
end management of
virtual
resources/devices/
machines.
29
Cloud Computing
Module 36

30
VM Migration
• VMs can be migrated
from one host to
another for:
1. Server Load
balancing
2. Server consolidation
3. Remedy for failover
hosts and VMs
4. Remedy for resource
shortage for a VM

31
VM Migration
• A VM can be in any of
the following states:
1. Powered-off
2. Suspended
3. Paused
4. Powered-on

32
VM Migration
• The following options
are available for VM
migration:
1. Cold migration: The
VM has to be
powered-off before
migration.
2. Warm migration:
Suspended VM
migration.
3. Live migration: For
powered-on VM with
zero downtime and
full availability.
33
VM Migration
• A VM is made of two
basic components:
• VM state: The
processor and RAM
contents
• Virtual hard disk:
Residing on network
storage or on host’s
hard disk
• Live migration of VM
means zero downtime
of OS, connectivity and
applications running on
the VM.
34
VM Migration
• For live migration, the
VM state is to be copied
from source to
destination host. The
virtual disk can also be
migrated through live
storage migration
feature of the
hypervisor.
• Modern day hypervisors
provide unbroken
network connectivity of
the VM during live
migration.

35
VM Migration
• During the live
migration, the state and
storage of the VM keeps
on working on source
host to avoid down-
time.
• For live migration of a
VM with the virtual hard
disk on network
accessed shared
storage, the virtual hard
disk need not to be
copied if the destination
host can access that
network based storage.
36
 VM Migration
• Migrating the virtual
hard disk is time
consuming as well as
network bandwidth
consuming due to multi-
Gigabyte migration.
• A better solution is to
use the network
storage.

end

37
Cloud Computing
Module 37

38
Virtualization for Data Center Automation
• A data center is a facility
with networked
computers and is used
by businesses and other
organizations to
process, store and share
large amounts of data.
• Companies like Google,
Yahoo, Amazon,
Microsoft, IBM, HP,
Apple etc. have invested
billions of dollars for
constrcting the data
centers.

39
Virtualization for Data Center Automation
• Data center automation
refers to the dynamic
provisioning of
hardware and software
resources to millions of
users simultaneously.
• Data centers can host
Clouds.
• Data center automation
is triggered by the
growth of virtualization
products.

40
Virtualization for Data Center Automation
• The data center owner
has three major
considerations:
1. Assuring
Performance and
QoS
2. Increase resource
utilization
3. Saving costs

41
Virtualization for Data Center Automation
• Enhanced resource
allocation (to jobs
and/or VMs) may be
performed in data
centers to assure
performance and QoS.

42
Virtualization for Data Center Automation
• The over allocation of
computing resources
may result in decrease
in average utilization of
these resources.
• This also leads to
increased costs due to
power consumption.

43
Virtualization for Data Center Automation
• Example: A VM hosted
on a server with 1.5 GHz
*4 cores and 16 GB of
RAM is allocated 1.5GHz
* 2 vCPUS, 4 GB vRAM
(half of the processing
and 1/4th RAM).

44
Virtualization for Data Center Automation
Suppose if there are two
such VMs. But the overall
average workload of the
hosted VMs keeps the
physical utilization to less
than 50%. This is a
resource wastage as 50%
of the resources remain
idle.

45
Virtualization for Data Center Automation
• Server consolidation is a
technique by which
more VMs are
aggregated on a single
server (by migrating
jobs/VMs to it) while
assuring performance
and QoS.
• This increases the
resource utilization
across data center.
• More servers are
available to take more
workload.
46
Virtualization for Data Center Automation
• More servers are
available to take more
workload. Otherwise,
the idle servers can be
shut down to save
power.
• Virtualization
technology also helps in
setting of virtual
storage (over VMs) to
offer virtual disks to
other VMs.

47
Virtualization for Data Center Automation
• Virtualization can
synchronize with cloud
management systems
to dynamically provision
cloud services and
billing systems.
• Hence, virtualization is
essential for Cloud
computing.

end

48
Cloud Computing
Module 38

49
Virtual Private Network (VPN)
• A VPN extends a private
network over public
network and enables
the users to
communicate as if their
devices are directly
connected to the
private network.
• A VPN creates a secured
and encrypted network
over a less secured
network such as the
Internet.

50
Virtual Private Network (VPN)
• Normally a VPN is
provided and managed
by a service provider.
• VPN allows the
corporate employees to
securely access the
applications hosted
over enterprise LAN.
• VPN is based upon IP
tunneling.

51
Virtual Private Network (VPN)
• IP tunneling or port
forwarding is the
transmission of private
network packets over a
public network
(Internet) as the
payload of public
network packets such
that the routing devices
do not come to know
about this.

52
Virtual Private Network (VPN)
• There are many
protocols for VPN
establishment and
encryption: IP Security
(IPSec), Secure Socket
Layer(SSL), Point-To-
Point Tunneling
Protocol (PPTP),
Multiprotocol Label
Switching (MPLS) etc.

53
Virtual Private Network (VPN)
• VPN although provide
secured connectivity to
extend a private
network but the
implementation may
have performance
issues.
• VPN is implementable
over Layer 1-3.

54
Virtual Private Network (VPN)
• Types of VPN:
1. Remote-access VPN:
A VPN client on
user’s device
connected to VPN
gateway of the
enterprise.
2. Site-to-site VPN:
Establishes a VPN
between two
networks over the
Internet by using
VPN gateway.

55
Virtual Private Network (VPN)
• VPN technology
provides access to cloud
resources. The VPN
gateway exists in the
cloud with a secure link
provided by the cloud
provider.

end

56
Cloud Computing
Module 39

57
Network Virtualization

58
 Network Virtualization

• Multiple virtual Network Interface Cards (vNIC)

are linked to physical NIC or pNIC through a
virtual Switch (vSwitch) inside a hypervisor.

VM 1 VM 2 VM 3

vNIC vNIC vNIC

vSwitch

pNIC

59
Network Virtualization
• A virtual network
consists of virtual nodes
and virtual links.
• Network virtualization
establishes the
coexistence of multiple
virtual networks.

60
Network Virtualization
• A virtual network
consists of virtual nodes
and virtual links.
• Network virtualization
establishes the
coexistence of multiple
virtual networks.

61
Network Virtualization
• Network virtualization
proposes the
decoupling of
traditional ISP
functionalities such as
infrastructure setup and
management from the
creation and
management of virtual
networks.

62
Network Virtualization
• It is possible to use
physical infrastructures
of multiple providers to
dynamically compose
virtual network/s.

63
Network Virtualization
• Technologies:
1. Virtual Local Area
Network (VLAN):
Logically grouping
the hosts with
common interest
into a single
broadcasting
domain.

64
Network Virtualization
• Technologies:
2. Virtual Private
Networks (VPN): A
dedicated
communications
network of
enterprise/s and user/s
by using tunneling
over public networks
(Internet).
end

65
Cloud Computing
Module 40

1
Networking Structure of Cloud Hosting Data center

• At the core of a cloud is

a cluster of VMs/
physical servers.
• There can be tens of
thousands of physical
servers in a data center.
• With each physical
server hosting multiple
VMs, the cloud hosting
networking structure
becomes a little
complicated.

2
Networking Structure of Cloud Hosting Data center

• The cluster nodes are

used for computations.
Some nodes are used
for workload allocation,
some for monitoring
and some for load
balancing.
• Some node called
gateway nodes provide
the interface of cloud
service/s to the outside
world (through
internet).

3
Networking Structure of Cloud Hosting Data center

• Cloud hosting data

center has a layered
architecture for the
Internet access.
• The servers are
physically connected to
layer 2 switches. There
is a top of rack (TOR) in
each rack. One server is
connected to only one
TOR switch.
• The TORs are connected
to aggregate switches
(AGS).
4
Networking Structure of Cloud Hosting Data center

• The AGSs provide the

cross rack inter VM
connectivity.
• There are a few access
routers (AR) and border
routers (BR) at layer 3.
• The layer 2 AGSs are
connected to BR
through AR.
• The BRs are connected
to Internet.

5
Networking Structure of Cloud Hosting Data center

• Some problems solved

by the hypervisor
solutions:
• The VMs hosted on
one server may belong
to different vLANs.
• A single vLAN may
span over multiple
data centers.
• A company may own
multiple data centers
and may want to
migrate the VMs
across data centers.
6
Networking Structure of Cloud Hosting Data center

Internet
BR BR

AR AR

AGS AGS

TOR TOR TOR

end

Rack Rack Rack

7
Cloud Computing
Module 41

8
Essential Characteristics of Cloud Computing
1. On-demand self-
service: The user can
automatically be
allocated the
computing resources
without any manual
operations (except
the initial signing up
process). The cloud
management
software handles the
resource management
and provisioning.

9
Essential Characteristics of Cloud Computing
2. Broad Network
Access: The cloud
resources can be
accessed through
network through
broad range of wired
and wireless devices.
Various connectivity
technologies are
available.

10
Essential Characteristics of Cloud Computing
3. Resource pooling:
Resources
(Computing, memory,
storage, network) are
available in volumes
and therefore can be
pooled. The resources
can be physical or
virtual. Multiple users
can simultaneously
share these resources
through dynamic
allocation and
reallocation.

11
Essential Characteristics of Cloud Computing
4. Rapid elasticity: The
cloud resources are
virtually unlimited. So
much so, the
provisioning of these
resources can shrink
and expand elastically
according to demand.

12
Essential Characteristics of Cloud Computing
5. Measured Service: The
resource usage is
charged by the
provider from users,
according to usage.

end

13
Cloud Computing
Module 42

14
Business Drivers for Cloud Computing
• Various business
drivers lure the
organizations to start
using Cloud.
• These include (but not
limited to):
• IT Capacity Planning
• Cost Reduction
• Organizational
Agility

15
Business Drivers for Cloud Computing
• IT Capacity Planning:
• It is the estimation
and fulfillment of
future IT requirements
of an organization.
• The over provisioning
of IT happens when
acquired equipment is
more than the
estimated
requirements.
Resulting in over
expenditure.

16
Business Drivers for Cloud Computing
• The under
provisioning occurs
when the equipment
turns out to be
inadequate to fulfill
the IT requirements
of the future.
• IT Capacity planning is
a difficult job as it
should cover the
fluctuating load.

17
Business Drivers for Cloud Computing
• Usually the companies
adopt any of the
following strategies:
• Lead Strategy: Adding
new IT capacity in
anticipation of future
needs.
• Lag Strategy: Adding
new IT capacity when
the IT resources reach
the full utilization
• Match Strategy:
Adding IT capacity in
small increments.
18
Business Drivers for Cloud Computing
• The capacity planning
may lead to adopting
the option of Cloud
Computing and then
planning for future
needs of Cloud
resources rental
instead of purchasing
the IT equipment.

19
Business Drivers for Cloud Computing
• Cost Reduction: The
costs include
• Cost of acquiring the
IT infrastructure
• Operational overheads
such as technical
personnel salaries,
upgrades, utility bills,
security, accounts and
administrative staff
salaries
• Why not choose the
Cloud instead ?

20
Business Drivers for Cloud Computing
• Organizational Agility:
It is the
responsiveness to the
change. We consider
the change in IT for
this topic.
• A possible shift,
upgrade or acquiring a
new software may
require to upgrade the
hardware.

21
Business Drivers for Cloud Computing
• The routine procedures
and the business may
come to halt or the
competitors may out
run if the organization
fails to invest in IT just
because of lack of
affordability.
end • The Cloud on the other
hand, just charges for
the usage of IT
resources, no need to
invest in infrastructure.

22
Cloud Computing
Module 43

23
Revisiting NIST Definition of Cloud Computing

• Cloud computing is a
model for enabling
ubiquitous, convenient,
on-demand network
access to a shared pool
of configurable
computing resources
(e.g., networks, servers,
storage, applications &
services) that can be
rapidly provisioned and
released with minimal
management effort or
service provider
interaction.
24
Revisiting NIST Definition of Cloud Computing

• Cloud computing can

fulfill the business
drivers such as
• IT Capacity Planning
• Cost Reduction
• Organizational
Agility
end

25
Cloud Computing
Module 44

26
Some key terms about Cloud Computing
• Some key terms and
concepts essential for
understanding Cloud
Computing course:
1. IT Resources
2. On-premises
3. Cloud Consumers
4. Cloud Providers

27
Some key terms about Cloud Computing
1. Cloud IT Resources: Can be
physical or virtual
resources (virtual
resources are implemented
in software):
• Physical/Virtual
machines/servers
• Physical/virtual storage

28
Some key terms about Cloud Computing
2. On-premises: An IT
resource which is
hosted/located at the
enterprise's premises.
• It is different from a
Cloud resource since a
Cloud resource is hosted
on Cloud.
• An on-premises IT
resource can be
connected to a Cloud
resource and/or can be
moved to a Cloud.
• However the distinction
is difficult for private
clouds.

29
Some key terms about Cloud Computing
3. Cloud Providers: The party
providing the cloud-based
IT resources.
4. Cloud Consumer: The user
of cloud-based IT
resources is called cloud
consumer.

end

30
Cloud Computing
Module 45

31
Scaling, Cloud Service Providers & Consumers

• Scaling: It refers to the

ability of an IT resource to
handle increased or
decreased usage demands.

32
Scaling, Cloud Service Providers & Consumers

• Following are the types of

scaling:
• Horizontal scaling: It is
the scaling out or scaling
in of the IT resources of
same type. The number
of resources increases or
decreases according to
load.
• Commodity hardware
can do the work,
instantly available IT
resources, not limited by
hardware capacity

33
 Scaling, Cloud Service Providers & Consumers

Physical server pool

VM VM VM VM VM VM
Virtual machines
(A) (A) (B) (A) (B) (C)

Horizontal scaling with increase in demand

34
Scaling, Cloud Service Providers & Consumers

• Vertical scaling: When an

IT resource is replaced
with a resource of higher
capacity (scaling up) or
when replaced with the
resource of lower
capacity (scaling down)
according to workload.
• Specialized server are
required, instantly
available IT resources,
additional setup is
required (downtime
required during
replacement), limited by
maximum hardware
capacity, less common in
Cloud.
35
Scaling, Cloud Service Providers & Consumers

Dedicated physical server with 4 CPUs

Vertical scaling with increase in demand

VM
Virtual machine with 2 vCPU

36
Scaling, Cloud Service Providers & Consumers

• Cloud Service:
Any IT resource
(software/VM) that is
made remotely available
by the cloud provider.
• Remember that not all the
IT resources deployed in a
cloud environment are
remotely accessible. Some
resources are used within
the Cloud for support and
monitoring etc.
• The human users interact
with a leased VM.

37
Scaling, Cloud Service Providers & Consumers

• Client programs interact

with cloud software
service/s through API calls.
• The software program and
service accessing a cloud
service is called a cloud
service consumer

end

38
Cloud Computing
Module 46

1
Benefits of Cloud Computing

• The immediate benefit

of using Cloud is the
reduction in initial cost.
• The initial costs include:
1. Infrastructure costs:
i. IT equipment
ii. Software
iii. Networking
iv. Construction costs
v. Installation costs

2
Benefits of Cloud Computing

• The infrastructure
costs can be regarded
as capital investments
or ownership costs.

3
Benefits of Cloud Computing

• The cloud saves the

initial upfront
ownership costs.
• The cloud offers
affordable and
attractive packages
for services obtained
in large volume.
• The cloud reduces
investment and
proportional costs.

4
Benefits of Cloud Computing

2. Proportional cost or
operational costs (as
discussed before):
The cloud rental can
replace this cost. The
rental costs are
highly competitive.

5
Benefits of Cloud Computing

• The cloud provider

can increase the
profit by increasing
the resource
utilization, using
proven practices
and by optimizing
the cloud
architecture.

6
Benefits of Cloud Computing

• Common measurable
benefits for the cloud
consumers are:
• Pay-as-you-go rental
for short term
usage
• The availability of
virtually unlimited
resources on
demand with
negligible wait time
for provisioning.

7
 Benefits of Cloud Computing

• The IT resources can

be added or
removed in a fine
grained level e.g., 1
GB of storage
increments
• Applications and
resources can be
migrated across
regions if required.
END

8
Cloud Computing
Module 47

9
Increased Scalability, Availability & Reliability

• Increased scalability:
The cloud can
dynamically and
instantly provide the
computing resources.
• This provision can be on
demand or as per user
configuration.

10
Increased Scalability, Availability & Reliability

• Similarly these IT
resources can be
released automatically
or manually with the
decrease in processing
demand.
• This dynamic scalability
avoids the over-
provisioning and under-
provisioning and the
associated
disadvantages.

11
Increased Scalability, Availability & Reliability

• Availability: The
availability of IT
resources sometimes
can be referred to profit
and customer retention.
• If an IT resource
becomes unavailable
(such as a database
dealing with clients’
orders) then this may
result in customer
dissatisfaction and loss
of business.

12
Increased Scalability, Availability & Reliability

• Reliability: The reliability

of IT resources is very
important for continual
business data
processing and
response time.
• The failure of any IT
resource can be cause
the collapse the IT
system. For example
failure of the Ethernet
switch may crash a
distributed application.

13
Increased Scalability, Availability & Reliability

• The modular structure

and resource
redundancy in cloud
increases the availability
and reliability. Cloud, on
the other hand provides
a guaranteed level of
availability and reliability
through a legal
agreement called
service level agreement
(SLA) between the
cloud provider and
cloud user.

14
Increased Scalability, Availability & Reliability

• The recovery time after

failure is the added
penalty. It is the time
when the system
remains unavailable.
• The modular structure
and resource
redundancy in cloud
increases the availability
and reliability. It also
improves the recovery
END time.

15
Cloud Computing
Module 48

16
Cloud Risks and Challenges

• The term vulnerability

refers to a state of
being attacked.
• Moving the business
data to cloud can
introduce vulnerabilities
and security risks.

17
Cloud Risks and Challenges

• The term security

framework refers to the
procedures and
practices for securing
the resource such as
data, network and IT
infrastructure.
• Unless the cloud
provider and cloud user
are covered under same
security framework, the
vulnerabilities are
unavoidable.

18
Cloud Risks and Challenges

• The cloud provider and

user have to be in a
trust relationship. The
factors affecting the
trust may include the
following facts:
• The data is being
accessed remotely.
• There are multiple
users sharing the cloud
based IT resources
such as virtual storage.

19
Cloud Risks and Challenges

• The cloud provider has

a privileged access to
the users’ data.
• The security of the
data depends upon
the security policies of
the provider and the
consumer.

20
 Cloud Risks and Challenges

• There can be malicious

consumers (human
and automated) who
can benefit from the
security vulnerabilities
of the cloud
environment by
stealing and/or
damaging the business
data.
END

21
Cloud Computing
Module 49

22
Cloud Risks and Challenges

• Reduced operational
governance control: The
cloud consumer gets a
lesser privileged control
over the resources
leased from the cloud.
• There can be risks
arising as to how the
cloud provider manages
the cloud.

23
Cloud Risks and Challenges

• An unreliable cloud
provider may not abide
by the guarantees
offered in SLA of the
cloud services. This will
directly affect the
quality of cloud
consumer solutions
(enterprise software)
which rely upon these
services.

24
Cloud Risks and Challenges

• The cloud consumer

should keep track of
actual level of service
being provided by the
cloud provider.
• The SLA violations can
lead to penalties
receivable from the
cloud provider.

25
Cloud Risks and Challenges

• Limited portability
between cloud
providers: Due to lack of
industry standards for
cloud computing, the
public clouds
environments remain
proprietary to their
providers.

26
Cloud Risks and Challenges

• It is quite challenging to
move a custom-built
software from one
cloud to another if it has
dependencies upon the
proprietary
environment (such as
security framework) of
the former cloud.

27
Cloud Risks and Challenges

• Multi-regional
compliance and legal
issues: Cloud providers
tend to set their data
centers in regions
favoring affordability
and/or convenient. This
may lead to legal issues
for cloud provider as
well as cloud
consumers.

28
Cloud Risks and Challenges

• Some countries such as

some UK laws require
the personal data of UK
citizens to be hosted
inside UK.
• Thus a cloud provider
with multi-regional data
centers including UK,
can not migrate the UK
citizen’s personal data
outside UK.

29
Cloud Risks and Challenges

• The UK citizen is legally

bound to keep the
personal data on clouds
hosted in UK only.

30
 Cloud Risks and Challenges

• Some countries such as

USA allows government
agencies’ access to data
hosted inside USA.
• Despite that the owners
of this data are neither
residing inside nor the
citizens of USA, but still
their data is accessible
by the USA government
agencies if hosted inside
END USA.

31
Cloud Computing
Module 50

32
Roles and Boundaries in Clouds

• Cloud provider: The

organization that
provides the IT
resources.
• Responsible for
providing IT resources
as per SLA.
• Also performs the
management and
administrative tasks to
assure flawless
provisioning of cloud
services.

33
Roles and Boundaries in Clouds

• A cloud provider usually

owns the IT resources
of the cloud.
• It is also possible that
the cloud provider
resells the cloud
services leased from
another cloud
providers.

34
Roles and Boundaries in Clouds

• Cloud consumer: The

organization or individual
who has contracted with
cloud provider to
lease/rent the cloud IT-
resources through user
interface and/or through
software API calls.
• In the later case, a cloud
consumer uses a cloud
service consumer (a
software program) to
interact/use a cloud
service.

35
Roles and Boundaries in Clouds

• Cloud Service Owner: Is

the one who owns the
cloud service. Can be:
• Cloud consumer: If the
deployed service is on
leased IT-resources.
• Cloud provider: If the
cloud provider has
deployed the service
on cloud IT-resources.
END • A cloud service owner
may not be the owner
of the cloud IT-resource.

36
Cloud Computing
Module 51

37
Roles and Boundaries in Clouds

• Cloud Resource
Administrator: This role
is responsible for
administering the cloud
resources (including
cloud services).

38
Roles and Boundaries in Clouds

• Cloud resource
administrator can be:
• Cloud consumer (as
cloud service owner)
• Cloud provider (when
the service resides
inside the cloud)
• Third party contracted
to administer a cloud
service

39
Roles and Boundaries in Clouds

• Additional roles:
• Cloud Auditor: Provides
an unbiased assessment
of trust building
features of the cloud.
These include the
security, privacy impact
and performance of the
cloud. The cloud
consumer may rely
upon the cloud audit
report for choosing a
cloud.

40
Roles and Boundaries in Clouds

• Cloud Broker: A party

that provides mediation
services to cloud
providers (seller) and
cloud consumers
(buyer) for the
purchasing of cloud
services.

41
Roles and Boundaries in Clouds

• Cloud Carrier: The party

responsible for
providing connectivity
between cloud provider
and cloud consumer.
The ISPs can be
assumed as cloud
carriers.
• The cloud provider and
cloud carrier are in legal
agreement (SLA) to
END assure a certain level of
connectivity and
network security.

42
Cloud Computing
Module 52

43
Roles and Boundaries in Clouds

• Organizational
boundary: This is a
boundary of ownership
ad governance of IT
assets of an
organization.
• Similarly, the cloud has
its organizational
boundary.

44
Roles and Boundaries in Clouds

• Organizational
boundary: This is a
boundary of ownership
ad governance of IT
assets of an
organization.
• Similarly, the cloud has
its organizational
boundary.

45
 Roles and Boundaries in Clouds

Organization A Cloud A

Cloud Cloud
Service Service
Consumer

Organizational boundary Organizational boundary

46
Roles and Boundaries in Clouds

• Trust boundary: When

an organization takes
the role of cloud
consumer, then it has to
extend its trust
boundary to include the
cloud resources.
• A trust boundary
represents a border
around trusted IT-
resources.

47
 Roles and Boundaries in Clouds

Trust boundary
Organization A Cloud A

Cloud Service Cloud

Consumer Service

Organizational boundary Organizational boundary

Cloud Computing
Module 53

49
Cloud Characteristics

• On-demand usage
• Ubiquitous access
• Multitenancy
• Resource pooling
• Elasticity
• Measured usage
• Resiliency

50
Cloud Characteristics

• Multitenancy: A
software architecture
software executing over
a server and serves
different users (tenants)
whereby each tenant is
isolated from the
others.
• Cloud computing serves
different cloud
consumers by using
virtualization software
frequently.

51
 Cloud Characteristics

Cloud Service Cloud Service

Consumer A Consumer B

Cloud Cloud
Service A Service B

Cloud Storage
Cloud Storage
Device A
Device B

In single-tenant environment, there is a separate IT resource for each tenant.

52
 Cloud Characteristics

Cloud Service Cloud Service

Consumer A Consumer B

Cloud Cloud
Service A Service B

Shared cloud
storage

Multi-tenant environment, a single instance of an IT resource such as Cloud

storage device serves multiple consumers.
53
Cloud Characteristics

• The cloud provider

pools the IT-resources
by using multitenancy
technology to
dynamically assign and
reassign these
resources according to
cloud consumers’
demands.

54
 Cloud Characteristics

• The physical as well as

virtual resources are
multi-tenanted (or
shared) by using
statistical multiplexing.

END

55
Cloud Computing
Module 54

56
Cloud Characteristics

• Resiliency: The ability of

a computer system to
recover from a failure is
called resiliency.
• The redundant
implementation of IT-
resources paves the
way to a resilient
system.

57
Cloud Characteristics

• The whole system is pre-

configured so that as
soon as a resource fails,
the processing is
automatically handed
over to the redundant
resource.

58
Cloud Characteristics

• Resiliency is one of the

features of cloud
computing whereby the
redundancy of IT-
resources is
implemented at
different physical
locations and/or in
different clouds.

59
Cloud Characteristics

• For example the data

can be kept at two
different locations and
replicated. If the
primary hard disk fails,
the secondary drive
takes the data
connectivity.

60
Cloud Characteristics

• A cloud service can be

configured at two
different VMs (A and B)
and each VM is placed
on a separate server or
a different cloud. VM B
is kept as failsafe
resource. In case VM A
fails, the VM B starts
processing the user
service user/s requests.

61
 Cloud Characteristics

Cloud Service Redundant implementation of

Consumer A the same service

Cloud Cloud
Service Service
A A

Resilient failover system

Cloud A Cloud B

62
 Cloud Characteristics

END

63
Cloud Computing
Module 55

1
IaaS, PaaS & SaaS Provisioning

• IaaS: The IT-resources

are typically virtualized
and packaged in a
preplanned way.
• The IT-resources are
usually freshly instanced
e.g., VMs.
IaaS, PaaS & SaaS Provisioning

• The cloud consumer has

a high level of control
and configuration-
responsibility.
• The cloud consumer
also has the duty of
configuring these
resources.
IaaS, PaaS & SaaS Provisioning

• Sometimes a cloud
provider will contract
IaaS offerings from
other cloud provider to
scale its own cloud
environment.
• The VMs can be
obtained specifying the
hardware requirements
such as processor
capacity, memory,
storage etc.
IaaS, PaaS & SaaS Provisioning

• PaaS: Delivers a
programming
environment containing
preconfigured tools to
support the
development lifecycle
of custom applications.
• PaaS products are
available with different
development stacks
such as Google App
Engine provides a
Python and Java
environment.
IaaS, PaaS & SaaS Provisioning

• The PaaS is chosen:

• To enhance or
substitute the on-
premises software
development
environment.
• To create a cloud
service in order to
provide a cloud service
to other cloud
consumers.
IaaS, PaaS & SaaS Provisioning

• The PaaS saves the

consumer from
administrative tasks
such as installations and
configurations to set up
the software
development
infrastructure.
• On the other hand the
cloud consumer has
lower level of control
over the underlying
infrastructure.
IaaS, PaaS & SaaS Provisioning

• SaaS: Is the software

hosted over cloud
infrastructure and
offered as a utility
services.
• SaaS is provided as a
reusable utility service
commercially available
to different users.
• A SaaS can be deployed
over IaaS and/or PaaS
instance. Whereby the
cloud consumer (of
IaaS/PaaS) becomes the
provider.
 IaaS, PaaS & SaaS Provisioning

• The service consumer

has a very limited
control over the
underlying SaaS
implementation.

end
Cloud Computing
Module 56

10
IaaS, PaaS & SaaS Comparison

• Control level:
• SaaS: Usage and usage
related configuration
• PaaS: Limited
administrative
• IaaS: Full
administrative
IaaS, PaaS & SaaS Comparison

• Functionality provided
to cloud consumer:
• SaaS: Access to front-
end user-interface
• PaaS: Moderate level
of administrative
control over
programming platform
• IaaS: Full
administrative control
over virtual resources
of the VMs
IaaS, PaaS & SaaS Comparison

• Common activities of
cloud consumer:
• SaaS: Use and
configure the service
• PaaS: Develop, debug
and deploy the cloud
services and cloud
based solutions
• IaaS: Installation and
configuration of
software, configure
the infrastructure of
VM
IaaS, PaaS & SaaS Comparison

• Common Cloud
Provider’s Activities:
• SaaS: Implementation,
management and
maintenance of cloud
service.
• PaaS: Providing the
pre-configured
programming
platform, middleware
and any other IT
resource needed.
 IaaS, PaaS & SaaS Comparison

• IaaS: Provisions and

manages the VMs and
underlying physical
infrastructure.
• The three cloud models
of cloud delivery can be
combined in a way that
one delivery model is
deployed over another.
end Such as:
• PaaS over IaaS
• SaaS over PaaS
• SaaS over PaaS over
IaaS
Cloud Computing
Module 57

16
Software as a Service (SaaS)

• NIST definition of SaaS:

“Software deployed as a
hosted service and
accessed over the
Internet.”
• The SaaS is a software
solution having the
code and data
executing and residing
on cloud.
• A user accesses the
SaaS through
browser.
Software as a Service (SaaS)

• Remember: The cloud

service consumer is a
temporary runtime role
assumed by a software
program when it
accesses a cloud service.
[Thomas Erl [2014], Cloud Computing
Concepts, Technology and Architecture,
Pearson]

• For the time being we

shall assume that the
browser acts as cloud
service consumer when
accessing a SaaS.
Software as a Service (SaaS)

• SaaS solutions eliminate

the need of on-premises
(data center based)
applications,
application
administration and data
storage.
• The customer is
allowed to adopt pay-
as-you-go type of
rental.
Software as a Service (SaaS)

• SaaS offers scalability

and device-
independent access to
the SaaS solution/s.
• SaaS provider assures
that the software
provided is solidly
tested and supported.
• The notable
disadvantage of SaaS
is that the data resides
off-premises.
Software as a Service (SaaS)

• The notable
disadvantage of SaaS is
that the data resides
off-premises.
• Therefore the data
security is of prime
importance because
the customers’ data
may be proprietary and
business-sensitive.
Software as a Service (SaaS)

• The SaaS provider

offers SaaS apps
executing over IT-
resources. These
resources can be from
a physical servers or a
VM owned/rented by
the provider.
• Each instance of a
SaaS app (consumed
by a user) is allocated
separate set of IT-
resources.
Software as a Service (SaaS)

• Classes of SaaS:
• Business logic:
Connect the
suppliers,
employees, investors
and customers.
• Example: Invoicing,
fund transfer,
inventory
management,
customer
relationship
management (CRM)
Software as a Service (SaaS)

• Collaboration:
Support teams of
people work
together.
• Examples: Calendar
systems, email,
screen sharing,
conference
management and
online gaming.
Software as a Service (SaaS)

• Office productivity:
Office environment
support.
• Examples: word
processors,
spreadsheets,
presentation and
database software.
Software as a Service (SaaS)

• Software tools: For

the support of
developing software
and solving
compatibility
problems.
• Examples: format
conversion tools,
security scanning,
compliance checking
and Web
development.
Software as a Service (SaaS)

• Software that are not

suitable for public
SaaS offerings
(according to NIST):
• Real-time software:
They require precise
response time. Due to
variable response time
and network delays,
these software are not
suitable to be offered
as SaaS. Such as flight
control systems and
factory robots etc.
Software as a Service (SaaS)

• Bulk-consumer data:
When extremely large
amount of data is
originating physically
at the consumer’s side
such as physical
monitoring and
patient monitoring
data. It is not feasible
to transfer this data in
real time over WAN to
SaaS provider.
 Software as a Service (SaaS)

• Critical software: A
software is labeled
critical if its failure or
delay in handling can
cause loss of life or loss
of property. These
software are not
suitable for SaaS
because achieving a
end continuous acceptable
reliability for critical
software in public SaaS
is quite challenging due
to (unreliable) public
network based access.
Cloud Computing
Module 58

30
SaaS Examples

1. Salesforce.com SaaS
for Customer
Relationship
Management (CRM)
• Manage sales
contacts and leads.
• Centralize the
contact.
information and
project details.
• The sales reports
from any place any
time.
SaaS Examples

• The sales reports

from any place any
time.
• Manages and
syncs sales
contacts and
meetings with
other tools such as
Microsoft Outlook.
SaaS Examples

2. Taleo SaaS for

Human Resources
Management
(HRM):
• Recruitment tools
to manage the
applicants’ data for
hiring purposes.
• Performance
management and
tracking tools for
employees’
evaluation.
SaaS Examples

• Performance
management and
tracking tools for
employees’
evaluation.
• Compensation
tools for rewarding
the employees
according to
performance.
• Workforce training
and professional
development tools
SaaS Examples

3. ADP SaaS for Payroll

Processing and
HRM:
• Cloud solution for
time management,
employees
benefits
calculation, worker
compensation and
HR issues.
SaaS Examples

4. Carbonite SaaS for

File Backups:
• Provides backup
services for precious
business data and
personal data. The
data is kept securely
and redundantly.
SaaS Examples

5. Microsoft Office 365

SaaS for Documnt
Creation, Editing and
Sharing:
• In order to provide
the documentation
tools at affordable
price and to compete
with the freeware
solutions, Microsift
offers its flagship
software suite on
monthly rental basis.
SaaS Examples

5. Microsoft Office 365

SaaS for Documnt
Creation, Editing and
Sharing:
The documents are saved
in cloud and are shareable
among multiple users.
Cloud Computing
Module 59

39
SaaS Software Stack

• The provider controls

most of the software
stack.

Application

Middleware

Operating
System

Hardware
SaaS Software Stack

• Application: Email
• Middleware: software
libraries, run time
environments (Java,
Python)

Application

Middleware

Operating
System

Hardware
SaaS Software Stack

• Service provider has

admin control over
application and total
control over the rest of
the layers.

Application

Middleware

Operating
System

Hardware
SaaS Software Stack

• Service consumer has

limited admin control
over the application
and no control over the
rest of the stack.

Application

Middleware

Operating
System

Hardware
 SaaS Software Stack

• A consumer can create,

send and manage the
emails and even the
email accounts.
• But the email provider
has absolute control
over the SaaS software
stack in order to
end perform its duties such
as provisioning,
management, updates
and billing in email app.
Cloud Computing
Module 60

45
SaaS Benefits

1. Modest software tool

footprint: There is no
need for complex
installation procedures
because the SaaS
applications and
accessible through
web browsers. This is
one of the reasons of
widespread use of
SaaS applications.
SaaS Benefits

2. Efficient use of
software licenses:
The license issuance
and management
procedure is quite
efficient. A single
client is issued a single
license for multiple
computers. This is
because the software
is running directly on
provider’s
infrastructure and
thus can be billed and
monitored directly.
SaaS Benefits

3. Centralized
management and
data: The consumer’s
data is stored in cloud.
The provider assures
the security and
availability of data. The
data seems centralized
for the consumer may
in fact be distributed
and replicated by the
provider. Data backup
is provided at possibly
additional charges.
SaaS Benefits

4. Platform
responsibilities
managed by providers:
Consumer does not has
to bother about
operating system type,
hardware and software
configurations,
software installation
and upgrades.
SaaS Benefits

5. Savings in up-front
costs: (as discussed
before) the up-front
costs such as
equipment acquisition
and hardware
provisioning etc. are
avoided by SaaS
consumer.
 SaaS Benefits

The provider is responsible

for operational issues such
as backups, system
maintenance, security
software, upgrades,
trouble shooting in
software, physical security
and hardware
management etc.
end
Cloud Computing
Module 61

1
SaaS: Issues and Concerns

• The NIST has identified

few issues and concerns
about SaaS.
• Most of these issues are
due to network
dependency of SaaS.
SaaS: Issues and Concerns

1. Browser based risks

and remedies: Since
the SaaS is accessed
through browser
installed on
consumers’ device,
the inherent
vulnerabilities of the
web browsers do have
impact over SaaS
security.
SaaS: Issues and Concerns

• Although the browsers

apply encryption upon
network traffic, yet
various network attacks
such as brute force and
man in the middle
attacks are possible upon
the SaaS data.
• The resources leased by a
consumer can be
hijacked by malicious
users due to poor
implementation of
cryptographic features
of browsers.
SaaS: Issues and Concerns

• If the consumer’s
browser is already
infected with a security
threat (due to a visit to
malicious website) then
later, the same browser
is used for SaaS access,
then the SaaS data might
get compromised.
SaaS: Issues and Concerns

• If a single consumer
accesses multiple SaaS
services using browser
instances, then the data
of these SaaS instances
may get mixed up.
SaaS: Issues and Concerns

• A few suggestions by
NIST:
• Use different browsers
to access each
different SaaS.
• Do not use the same
web browser for web
surfing and SaaS
access.
• Use a VM to access the
SaaS.
SaaS: Issues and Concerns

2. Network dependence:
SaaS application
depends upon reliable
and continuously
available network.
• The reliability of a
public network
(Internet) can not be
guaranteed as
compared to dedicated
and protected
communication links of
private SaaS
applications.
SaaS: Issues and Concerns

3. Lack of portability
between SaaS clouds:,
It may not be trivial to
import export data
among different SaaS
applications deployed
over different clouds
due to customized
development and
deployment of SaaS
applications and data
formats.
SaaS: Issues and Concerns

4. Isolation vs. Efficiency

(Security vs. Cost
Tradeoffs): The SaaS
provider has to make
a trade-off decision as
to deploy separate IT-
resources (such as
VMs) for each client or
concurrently server
multiple clients
through a single
deployment of SaaS
application.
Cloud Computing
Module 62

11
NIST Recommendations for SaaS

1. Data protection: The

consumer should
analyze the data
protection,
configuration, database
transaction processing
technologies of SaaS
provider. Compare
them with the
confidentiality,
integrity, availability
and compliance
requirement of the
consumer.
NIST Recommendations for SaaS

2. Client
device/application
protection: The
consumer’s client
device (browser
running over a
computer) should be
protected to control
the exposure to
attacks.
NIST Recommendations for SaaS

3. Encryption: Strong
encryption algorithm
with key of required
strength should be
used for each web
session as well as for
data.
 NIST Recommendations for SaaS

4. Secure data deletion:

The data deletion
through consumer’s
request should be
reliably done.

end
Cloud Computing
Module 63

16
PaaS Overview

• According to NIST,
PaaS provides a toolkit
for conveniently
developing, deploying
and administering
application software
which can support a
large number of users,
process large volumes
of data and can be
accessed over
Internet.
PaaS Overview

• What does PaaS

clouds really provide:
a set of software
building blocks, a set
of development tools
(languages and
compilers) and
supporting
environments for run-
time of applications
developed over PaaS.
PaaS Overview

• PaaS clouds also

provide tools to
deploy the developed
applications.
• Additionally, the PaaS
clouds provide
processing, storage
and networking
resources.
PaaS Overview

• PaaS consumers:
• Application developers
• Application testers
• Application deployers
• Application
administrators
• Application end users
(SaaS users)
• The consumers are
charged according to
tools and IT-resources
usage.
PaaS Overview

• PaaS Software stack:

The cloud provider
fully controls the
hardware and OS
layers.

Application

Middleware

Operating
System

Hardware
PaaS Overview

• PaaS Provider/
Consumer Scope of
Control: The provider
has administrative
control of middleware.

Application

Middleware

Operating
System

Hardware
PaaS Overview

• The provider has no

control over
application layer.

Application

Middleware

Operating
System

Hardware
PaaS Overview

• Remember that the

application developed
by using PaaS is
deployed as SaaS and
the PaaS consumer
has full administrative
control over that SaaS.
• The provider however
controls the runtime-
environment which is
necessary for PaaS
application.
PaaS Overview

• PaaS billing: Usually

based on:
• Number of consumers
• kind of consumers
(e.g., developers vs.
application end users)
• Storage, processing,
or network resources
consumed by the
platform
• Requests serviced
• The time the platform
is in use.
 PaaS Overview

• SaaS billing: Based on

• Number of users
• Time in use
• Per-execution, per-
record-processed
• Network bandwidth
consumed
end
• Quantity/duration of
data stored
Cloud Computing
Module 64

27
PaaS Examples

• We are going to
discuss a few
examples of PaaS.
PaaS Examples

1. Google App Engine

(GAE): Allows the
users to create and
host web based (Java,
Python & Go)
applications running
over the infrastructure
and services provided
by Google.
• GAE is a free service
until the application
grows to a significant
size.
PaaS Examples

2. Force.com as a PaaS:
This is a service of
Salesforce.com (a
SaaS provider). It
offers four different
programming
environments for
nonprogrammers,
programmers and
software vendors.
PaaS Examples

• Nonprogrammers can
create finance, HR etc.
applications and
websites without
coding by using drag
drop of controls.
PaaS Examples

• Programmers can
develop Java
applications and deploy
them as SaaS.
• The software vendors
can distribute and
update their
applications over cloud
by using Force.com.
PaaS Examples

3. LongJump as a PaaS:
Supports the entire
cycle of software
development from
requirement gathering
to application release
and support. It is free
of cost.
PaaS Examples

3. Openshift as a PaaS: It
is a PaaS offering from
Red Hat which is also
the distributor for Red
Hat Linux. Openshift
PaaS provides the
primary development
tools for cloud based
solutions written in
PHP, Python and
Ruby.
PaaS Examples

• Openshift also
provides development
tools for Linux-based
solutions written in C
programming
language.
 PaaS Examples

4. Windows Azure and

SQL Azure as a PaaS:
Provided by Microsoft
as a paid service. The
users can develop
applications in .Net as
well as Java, PHP and
Ruby.
end
• SQL Azure provides
database solutions for
application developed
and running inside
Windows Azure.
Cloud Computing
Module 65

37
Benefits and Disadvantages of PaaS Solutions

• Benefits:
i. Lower total cost of
ownership in terms of
hardware and software
investment.
ii. Lower administrative
overhead of system
development.
iii. No requirement of
software upgrades of
tools.
iv. Faster application
development and
deployment.
Benefits and Disadvantages of PaaS Solutions

v. Scalable resources
available for the
applications. The user
pays only for the
resources used.
Benefits and Disadvantages of PaaS Solutions

• Disadvantages:
i. The inherent problem
of data placed offsite
raises the security
concerns.
ii. The integration of
PaaS applications
with on-site legacy
solutions is not trivial.
iii. The PaaS provider has
to be trusted for data
and application
security.
Benefits and Disadvantages of PaaS Solutions

iv. The issues of SaaS are

also the issues of
PaaS such as browser
based risks, network
dependence and
isolation vs efficiency.
v. Portability of PaaS
applications across
end
different providers
may not be possible
sue to incompatibility
in coding structures
(hash, queue, file
etc.).
Cloud Computing
Module 66

42
PaaS Recommendations

1. Generic interfaces:
The consumer should
make sure that the
interfaces for hash
tables, queues and
files etc. are generic
so that there will be
less issues of
portability (among
PaaS providers) and
interoperability (of
applications) in future.
PaaS Recommendations

2. Standard language
and tools: Choose a
PaaS provider which
offers standardized
language and tools
unless it is absolutely
unavoidable to use the
proprietary languages
and tools.
PaaS Recommendations

3. Data access: The

provider with the
standardized data
access protocol (such
as SQL) should be
preferred.
PaaS Recommendations

4. Data protection: The

confidentiality,
compliance, integrity
and availability needs
of the organization
should be compared
with the data
protection
mechanisms of the
provider.
PaaS Recommendations

5. Application
framework: The PaaS
providers which offer
the features in
application
development
framework for
eliminating security
vulnerabilities of the
application should be
chosen.
PaaS Recommendations

6. Component testing:
The software libraries
provided by the PaaS
provider should be
aiming at providing
proper functionality
and performance.
 PaaS Recommendations

6. Security and secure

data deletion: Ensure
that the PaaS
applications can be
configured to run in a
secure manner (e.g.,
using cryptography
during
communication) and
end
that a reliable
mechanism for data
deletion is provided by
the PaaS provider.
Cloud Computing
Module 67

50
IaaS Overview

• As an alternative to
PaaS, some
consumers may prefer
to use IaaS in order to
have management
control over the IT
resources.
• The IaaS provider
makes available the
computing resources
in the form of VMs.
• The consumer has the
duty of installing OS
and software.
IaaS Overview

• The provider also

provides stable
network access,
network components
such as firewalls, and
data storage.
PaaS Overview

• IaaS
Provider/Consumer
Scope of Control: The
provider has no
control over top three
layers.
Application

Middleware

Operating
System

Hypervisor

Hardware
PaaS Overview

• IaaS
Provider/Consumer
Scope of Control: The
provider has admin
control over
hypervisor and total
control over hardware
layer.
PaaS Overview

• IaaS
Provider/Consumer
Scope of Control: The
consumer has total
control over top three
layers.
Application

Middleware

Operating
System

Hypervisor

Hardware
PaaS Overview

• IaaS
Provider/Consumer
Scope of Control: The
consumer can request
the provider to deliver
a VM from hypervisor
layer.
• The consumer has no
control over hardware
layer.
 IaaS Overview

• Customer billing:
• Per CPU hour
• Data GB stored per
hour
• Network bandwidth
consumed, network
infrastructure used
(e.g., IP addresses) per
end hour
• Value-added services
used (e.g., monitoring,
automatic scaling).
Cloud Computing
Module 68

58
IaaS Operational Overview

• The operational
infrastructure of IaaS
will be discussed.
 IaaS Operational Overview

Legend
Cloud
DOS DOS – Data Object Storage
Manager PLS – Persistent Local Storage
Network WAN
LAN

Cluster Cluster PLS

PLS
Manager Manager

Queries, Cmds
Queries, Cmds

reports
reports

Computer Manager Computer Manager

CMn
…
CM1
…
CMn
…
CM1
(CMj ) … (CMj )

hypervisor hypervisor

VM1 VM2
… VMn VM1 VM2
… VMn
Cloud Computing
Module 68

61
IaaS Benefits

• Saving in upfront cost:

As in SaaS and PaaS.
Although the
responsibility of
installing OS and
software is of the
consumer.
IaaS Benefits

• Full administrative
control over VM:
• Start, shut down,
pause
• Installation of OS
and applications
• Accessing VM
through network
services of VM
through a network
protocol such as
Secure Shell.
IaaS Benefits

• Flexible and scalable

renting: The VMs can
be rented in any
volume desired by the
consumer. The rental
for each VM can be on
usage (of raw
resources such as
CPU, memory,
bandwidth, storage,
firewall, database
etc.) basis.
 IaaS Benefits

• Portability and
interoperability with
legacy applications:
Since the consumer has
full control over the VM
to install OS and other
applications, the legacy
applications (which are
usually installed on
end consumer owned
server/s) can be
configured to run with
or ported to the VM.
Cloud Computing
Module 70

66
IaaS Issues and Concerns

• Network dependence
• Browser based risks
• As discussed for SaaS
and PaaS.
IaaS Issues and Concerns

• Compatibility with
legacy software
vulnerabilities: Since
the consumer is
allowed to install the
legacy applications on
VMs rented through
IaaS, this exposes the
VMs to the
vulnerabilities in those
legacy software.
IaaS Issues and Concerns

• Implementation
challenges exist for
VM isolation: In order
to prevent the VMs
from eavesdropping
other VMs mounted
over same server, the
isolation features of
hypervisor are utilized.
But these features
may not withstand a
sophisticated attacks.
IaaS Issues and Concerns

• Dynamic network
configuration for VM
traffic isolation: A
dynamic network path
is provided from VM
to consumer when a
VM is rented. The
provider has to isolate
VM consumers from
accessing the network
traffic of other
consumers.
IaaS Issues and Concerns

• Data erase practices:

When a VM is no
longer rented by a
consumer, the virtual
drive of that VM must
be erased/overwritten
multiple times to
eliminate any chance
of residual data access
by the next consumer
of that VM.
 IaaS Issues and Concerns

• NIST
recommendations for
IaaS: The provider
should implement
data and network
traffic isolation for the
VM consumers. The
features of data
security as well as
end
secure deletion of
residual data of VM
consumer.
Cloud Computing
Module 71

73
Identity as a Service (IDaaS)

• Today within most

companies, the users
may have to log in to
several applications
servers (on premises
and/or cloud) to
perform daily tasks.
Some of these
systems may be cloud
based.
• The user has to
remember multiple
logins and passwords.
Identity as a Service (IDaaS)

• When a user leaves a

company, the related
logins and passwords
must be deleted.
• The identity
management is a
complex task and
therefore provided as
a service for cloud
consumers.
• For example single
sign on (SSO).
Identity as a Service (IDaaS)

• Single sign on (SSO)

software is installed
over authentication
server.
• Before connecting to
application servers,
the user connects with
the authentication
server to obtain a
secure ticket.
 Identity as a Service (IDaaS)

User

Payroll server
HR server

Training
Authentication server
server
Identity as a Service (IDaaS)

• The authentication
server maintains the
user login security
credentials required
by application servers.
• When the user leaves
the company, only the
user’s login on
authentication server
is needed to be
disabled to block the
user’s access to all the
application servers.
 Identity as a Service (IDaaS)

• There are a few

examples of IDaaS
providers for on-
premises and cloud
applications such as
Ping IDaaS and
PasswordBank IDaaS.

end
Cloud Computing
Module 72

1
IDaaS: OpenID

• It is a popular example
of Identity as a
Service(IDaaS).
• Allows the users to sign-
in to multiple websites
by using a single
account.
• Solves a lot of problems
related to multiple log-
in accounts per user.
IDaaS: OpenID

• Why use OpenID:

• Avoid too many user
names and passwords.
• Overcoming the
scarcity of desired user
names.
• Account management
is difficult otherwise.
• Avoid filling long forms
for creating logins
again and again.
IDaaS: OpenID

• OpenID is not controlled

by any organization
and/or person.
• There are a number of
companies (providers)
which provide OpenID
accounts. These include:
Google, Microsoft,
Yahoo, Amazon,
SalesForce etc.
• There are more than 1
billion OpenID accounts
which are accepted by
over 50,000 websites.
IDaaS: OpenID

• How does it work:

• A user creates an
OpenID login through
a suitable provider.
• The user visits a
website which is
compatible with
OpenID.
• The (visited) site
prompts the user to
sign-in with the
OpenID credentials.
IDaaS: OpenID

• How does it work:

• The user is redirected
to the OpenID
provider’s website.
• The user opts to share
the credentials/token
with the (visited)
website.
• The user provides
login and password at
the OpenID provider’s
website.
 IDaaS: OpenID

• How does it work:

• If the user is verified,
the OpenID provider
confirms the (visited)
website.
• The user is redirected
to the (visited)
website which accepts
end the user as
authenticated user.
Cloud Computing
Module 73

8
Data Storage in Clouds: Network Storage

• Computers attached to
a local area network
(LAN) may require
additional storage space
to support file sharing,
file replication and
storage for large files.
• Traditionally this
additional space is
provided through file
servers which have
larger disk capacity.
Data Storage in Clouds: Network Storage

• With the evolution of

computer networks, the
file server was extended
through the use of
storage area network
(SAN).
• The SAN enabled storage
devices are attached to
the network.
• The software running
over SAN devices allows
direct access to these
devices throughout
network.
Data Storage in Clouds: Network Storage

• Later on, a class of

storage devices
emerged to be
implemented as
network attached
storage (NAS).
Data Storage in Clouds: Network Storage

• Advantages of network
storage (particularly of
SAN) are:
• Data reliability and
reconstruction
through replication.
• Better performance
than file server.
end
• Compatibility with
common file systems
and operating
systems.
• Best choice for
backups.
Cloud Computing
Module 74

13
Data Storage in Clouds: Cloud Based Data Storage

• Cloud storage is the

next step in the
evolution of network
storage devices.
• Instead of storing the
data locally, the data
can be stored on cloud
and can be accessed
through web.
• The user can have
virtually unlimited
storage space available
at affordable rates.
Data Storage in Clouds: Cloud Based Data Storage

• There are various modes

of data access in Cloud:
• Using web browser
interfaces to move the
files to and from the
cloud storage.
Data Storage in Clouds: Cloud Based Data Storage

• Through a mounted
disk drive that appears
local to the user’s
computer.
• Through API calls to
access the cloud
storage.
Data Storage in Clouds: Cloud Based Data Storage

• There are a number of

cloud storage providers
which offer file storage,
sharing and
synchronization. Such
as:
• Carbonite
• pCloud
end
• Dropbox
• ElephantDrive
• These providers offer a
certain volume of free
storage as well as paid
storage at low prices.
Cloud Computing
Module 75

18
Data Storage in Clouds: Advantages & Disadvantages

• Advantages:
• Scalability: The user
can scale the storage
capacity (up or down)
according to
requirement.
• Various convenient
costing models are
available from one
time payment to
monthly payment to
pay as per use.
Data Storage in Clouds: Advantages & Disadvantages

• Advantages:
• Reliability: The storage
providers provide the
assurance for data
reliability (through
replication).
• The data can be
accessed worldwide
by using Internet.
• Various methods of
data access are
available (as discussed
before).
Data Storage in Clouds: Advantages & Disadvantages

• Disadvantages:
• Performance: Because
of the Internet based
access, the cloud
storage can never be
as fast as SAN or NAS
based local storage.
• Security: Not all the
users may be able to
trust the cloud
provider for the users’
data.
Data Storage in Clouds: Advantages & Disadvantages

• Disadvantages:
• Data orphans: The user
has to trust the data
deletion policies of the
provider. The files (on
cloud storage) deleted
by the user may not be
immediately (or ever)
end be deleted from the
cloud storage.
Cloud Computing
Module 76

23
Data Storage in Clouds: Cloud Based Backup Systems

• The term backup refers

to the copying of (data
and/or database) files to
a secondary site for
preservation in case of
device or software
failures.
• Backup is an important
part of disaster recovery
plan.
• In case of a disaster, the
data can be restored to
the state of last backup.
Data Storage in Clouds: Cloud Based Backup Systems

• Cloud based backup

system comprises of
procedures to send the
copy of data over a
proprietary or public
network to a remote
server hosted by the
cloud service provider.
Data Storage in Clouds: Cloud Based Backup Systems

• The provider charges

the user according to
number of accesses or
data volume or number
of users.
Data Storage in Clouds: Cloud Based Backup Systems

• Cloud based backup or

online backup system is
implemented through a
client software installed
on the user’s computer.
The software collects,
compresses and sends
the data to cloud
backup on timely basis.
Data Storage in Clouds: Cloud Based Backup Systems

• Advantages:
• The data is backed up
in encrypted form.
• Backup can be
performed on the
convenience of user
(daily, weekly,
monthly).
• The user can easily
retrieve the backup
files from the cloud.
Data Storage in Clouds: Cloud Based Backup Systems

• Disadvantages /
Limitations:
• Due to security
concerns, the critical
data backup is
preferably stored on
local storage.
• The long term data
end
storage in heavy volume
over cloud may have
humongous cost.
• Due to network cost,
the incremental backup
is preferred.
Cloud Computing
Module 77

30
Data Storage in Clouds: Database and Block Storage

• A Cloud database is a
database that resides on
Cloud platform.
• The Cloud database can
be accessed by:
• The applications
hosted on Cloud
• The application
hosted locally (can
access through
Internet)
Data Storage in Clouds: Database and Block Storage

• The cloud database is

provisioned in either of
the following methods:
• Installed on a rented
VM by the user
• As part of PaaS
• Provided as a service
by cloud provider or
the database
companies.
Data Storage in Clouds: Database and Block Storage

• Advantages of Cloud
based Database
solutions:
• Cost effective
scalability as per use
• High availability of
database software
through redundant
hardware (minimizes
the downtime in case
of failure)
Data Storage in Clouds: Database and Block Storage

• High availability of
data due to replication
of database
• Reduced
administration of
database provided as
service or as part of
PaaS.
Data Storage in Clouds: Database and Block Storage

• Disadvantages of Cloud
based database
solutions:
• The user may not trust
the cloud provider
regarding sensitive
data
• Due to Internet based
access, the Cloud
based database is not
as fast as a locally
installed database.
Data Storage in Clouds: Database and Block Storage

• There are a number of

cloud based database
providers such as:
• Oracle
• Amazon
• Microsoft
Data Storage in Clouds: Database and Block Storage

• Cloud based block

storage is a sequence of
bits and provided as a
block on cloud storage.
• It is suitable in the
following situations:
• When the data may
not map properly on a
file system or on a
database
• The application
developer wants to
store data in a
customized file system
Data Storage in Clouds: Database and Block Storage

• Amazon Elastic Block

Store (EBS) is a highly
available, scalable and
reliable block storage
solution which supports
block sizes of up to 1
terabytes.

end
Cloud Computing
Module 78

39
Collaboration in the Cloud: Background

• Collaboration is defined
as the process in which
two or more people
work together to
achieve a goal.
• Traditionally, the
collaboration has been
achieved through face
to face meetings in
conference rooms.
Collaboration in the Cloud: Background

• Some team members

had to travel (from near
or far) to attend the
meetings.
• Those who could not
personally arrive at the
meeting had either of
the following two
choices:
• Phone call to a speaker
phone placed at the
conference table
• Study the minutes of
meeting
Collaboration in the Cloud: Background

• A solution that could

reduce the
requirements of
personal meetings was
required to save time
and effort and to
increase the
productivity from the
collaborations.
Collaboration in the Cloud: Background

• The web based

collaboration began
with the web mail.
• Users can compose,
send, receive and read
the emails by using the
web browser and
Internet connection.
• A single user can
address multiple
recipients in a single
mail.
Collaboration in the Cloud: Background

• The instant messaging

(IM) provide a real time
exchange of messages
and replies (chat) by
using messaging
software.
• IM is another form of
traditional
collaboration. Current
tools for IM allow file
exchange and
audio/video calling.
Collaboration in the Cloud: Background

• Voice over Internet

Protocol (VoIP) enables
the users to make
phone calls over the
Internet.
• VoIP tools such as Skype
provide a convenient
way to perform
end conference calls by
using computers and
mobile phones.
Cloud Computing
Module 79

46
Cloud based Phone & Fax Systems

• Sending and/or
receiving fax
traditionally required
the fax machine and
telephone connection.
• Similarly, phone calling
has been dependent
upon telephone
infrastructure.
Cloud based Phone and Fax Systems

• In modern days, many

companies have started
providing cloud based
calling and cloud based
fax services.
• These companies have
all the calling/fax
operations performed
over the Cloud and
provisioned over the
Internet.
Cloud based Phone and Fax Systems

• Taking example of
Google Voice Phone
System: The account
holder receives the
services of call
answering and voice
mail.
• The user can even
configure the service to
forward the incoming
phone calls to a cell
number.
Cloud based Phone and Fax Systems

• Google delivers the

voice messages left by
the callers as audio
messages as well as in
the form of text which
are receivable anywhere
through the Internet.
Cloud based Phone and Fax Systems

• Cloud based fax service

provided by various
companies is
provisioned as a
separate virtual number
to each subscriber. This
number corresponds o a
virtual fa machine.
• The fax received over
the virtual fax machine
are delivered through
email as PDF
attachment.
Cloud based Phone and Fax Systems

• Similarly, to send a fax,

a simple email (with PDF
file) to virtual fax
account will send the
fax to recipient/s.

end
Cloud Computing
Module 80

53
Editing the Shared Files in Cloud

• As we have seen that

data and files can be
stored on Cloud
storage.
• It is also possible to edit
the files (located on
Cloud storage) shared
among concurrent
users.
• Provides another way of
collaboration.
Editing the Shared Files in Cloud

• A number of service
providers offer the
editing of shared files
such as text,
spreadsheet and
presentation files. These
include the famous
providers:
• Dropbox
• Microsoft
• Google
Editing the Shared Files in Cloud

• Dropbox offers file

sharing through public
folders among the
Dropbox users.
• It is allowed to edit the
MSWord, Excel and
PowerPoint files in
browser and without
the MSOffice installed.
• Simultaneous users can
edit a shared document.
: Editing the Shared Files in Cloud

• Google provided Google

Docs service offers web-
based free access to a
word processor,
spreadsheet and
presentation programs
to create, share, edit,
print and download the
end documents stored on
Cloud.
• Google Docs can be
shared through simple
email link.
Cloud Computing
Module 81

58
Collaboration in the Cloud: Collaborative Meetings

• Collaborative meeting
can be performed by
using the software
hosted on Cloud.
• Organizations get a cost
effective virtual meeting
as an alternative to face
to face meetings.
Collaboration in the Cloud: Collaborative Meetings

• The features of cloud

based collaborative
meetings are:
• Streaming video to
allow face to face
interaction
• Shared whiteboards to
control the
presentation
• Shared applications to
demonstrate software
in live environment
• Meeting recordings for
playback and sharing
Collaboration in the Cloud: Collaborative Meetings

• GoToMeeting is one of
the leading providers of
virtual meetings.
• Can support face to face
meetings and web
seminars (webinars)
with more than 1000
attendees.
end
• The video recording of
virtual meetings and
webinars can also be
used for virtual training
and reference purposes
as well.
Cloud Computing
Module 82

1
Collaboration by Social Media & Video Streaming

• Social media and

streaming video
contents provide yet
another way for
collaboration.
• Cloud hosted social
media such as Facebook
and SalesForce.com’s
Chatter tool are
available for
collaboration among
team members.
Collaboration by Social Media & Video Streaming

• The team member can

easily exchange
updates, comments and
reviews regarding
different tasks.
• Files can be shared
among the team
members.
Collaboration by Social Media & Video Streaming

• Photos and videos can

be uploaded and shared
to demonstrate a
situation.
• Live video streaming
can also be broadcasted
if required.
Collaboration by Social Media & Video Streaming

• YouTube offers a free,

reliable and Web
accessed cloud storage
for video contents
worldwide.
• Videos created for
collaboration can be
shared among team
members and publicly
as well.
Collaboration by Social Media & Video Streaming

• The collaborative videos

may include technical
training clips,
discussions and/or site
coverage etc.
• The viewers can discuss
and upload written
comments on the video
end clip.
Cloud Computing
Module 83

7
Cloud Deployment Models: Public Cloud

• Public cloud is one of

the deployment models
of Cloud through which
the IT resources are
publicly available and
accessible trough public
Internet.
Cloud Deployment Models: Public Cloud

• Characteristics of Public
Cloud according to NIST:
• The consumer is
generally not aware of
the location of IT
resources unless a
location restriction is
imposed by either of
provider or consumer.
Still it is difficult for the
consumer to verify the
location on map from
where the IT resources
are being provisioned.
Cloud Deployment Models: Public Cloud

• Characteristics of Public
Cloud according to NIST:
• The consumer
workload may be a co-
resident of the
workload of other
consumer (multi-
tenancy) which may
include the rivals,
adversaries and in
worst case, the
attackers.
Cloud Deployment Models: Public Cloud

• Characteristics of Public
Cloud according to NIST:
• The consumer has
limited visibility of the
software and
procedures of the
provider. The
consumer has to trust
the provider for
securing the
consumer’s data and
fully disposing the
deleted data.
Cloud Deployment Models: Public Cloud

• Characteristics of Public
Cloud according to NIST:
• The consumer
undergoes a limited
upfront cost regarding
the provisioning of IT
resources as compared
to in house or locally
setting up the IT
infrastructure.
Cloud Deployment Models: Public Cloud

• Characteristics of Public
Cloud according to NIST:
• Thanks to the
workload
management, dynamic
collaboration among
cloud providers and
(generally) large
setups, the public
clouds can give the
illusion of unlimited
resources and
elasticity to the
consumers.
Cloud Deployment Models: Public Cloud

• Characteristics of Public
Cloud according to NIST:
• The provider is in a
limited legal Service
Level Agreement (SLA)
with the consumer.
The SLA covers the
minimum performance
end assurance/s by the
provider and penalty in
case of violation to the
assurance/s.
Cloud Computing
Module 84

15
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• The cloud
infrastructure is
provisioned for
exclusive use by a
single organization
comprising multiple
consumers (e.g.,
business units).
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• It may be owned,
managed, and
operated by the
organization, a third
party, or some
combination of them,
and it may exist on or
off premises.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• The private cloud users
depend upon the local
area network if the
cloud is locally
deployed and
accessed from a single
site.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• For multi-site access
and outsourcing, the
dedicated leased
secure communication
lines should be used.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• Consumers are needed
to be trained for
working in Cloud
environment.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• Consumers have no
knowledge of the
location of their
workload. Even in on-
site deployment, a
consumer can not
pinpoint a server for
the location of
workload.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• However, in case of
outsourced Private
Cloud, the consumer
organization may have
some knowledge of
the cluster location
and network segment
serving the Private
Cloud at the provider’s
end.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• Consumer workload is
vulnerable to cons of
multi-tenancy from the
insider malicious
colleagues.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• Modest cost for
outsourced private
Cloud (excludes
infrastructure cost):
Negotiation with the
provider, Upgradation
in network equipment,
updating of legacy
software to work on
Cloud, training of staff
etc.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• Significant cost for
onsite private Cloud
(includes the data
center and
infrastructure cost):
Updating of legacy
software to work on
Cloud, training of staff
etc.
Cloud Deployment Models: Private Cloud

• Characteristics of
Private Cloud according
to NIST:
• Resource limitation in
on-site private Cloud
but extendible
resources available in
case of outsourced
end private Cloud.
Cloud Computing
Module 85

27
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• The cloud
infrastructure is
provisioned for
exclusive use by a
specific community of
consumers from
organizations that
have shared concerns
(e.g., mission, security
requirements, policy,
and compliance
considerations).
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• It may be owned,
managed, and
operated by one or
more of the
organizations in the
community, a third
party, or some
combination of them,
and it may exist on or
off premises.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• For the onsite
Community Cloud, the
resource sharing
among the
participating
organizations has to
be decided explicitly or
implicitly.
• At least one member
of the community
should provide Cloud
services.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• Network dependency:
In case of on-site
deployment, the
network dependency
is similar to on-site
distributed Private
Cloud setup. The
performance and
security can be
enhanced through
dedicated secured
communication lines.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• Network dependency:
The members can also
use encryption over
Internet for the
network access to the
Community Cloud
resources.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• IT skills are required to
manage the
Community Cloud
deployment and
operations in both the
participants (providing
Cloud services) and
consumer members of
the community.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• Workload locations are
generally hidden from
the community
members unless a
participant member
decides to outsource
the Cloud services
(similar to outsourced
Private Cloud). In this
case, prior approval
and documentation
should take place.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• Multi-tenancy cons are
similar to onsite
Private Cloud scenario.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• The upfront cost for
consumer-only
member is same as of
outsourced Private
Cloud. While for
participant members
(onsite deployment),
the upfront cost is
similar to onsite
Private Cloud.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• The onsite deployment
of Community cloud
suffers from resource
shortage as of onsite
Private Cloud because
each participant
organization has
limited resources.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• Extensive resources
are available for
outsourced
Community Cloud just
like outsourced Private
Cloud.
Cloud Deployment Models: Community Cloud

• Characteristics of
Community Cloud
according to NIST:
• Due to a number of
members, there are a
number of security
perimeters (hence
complex
end cryptography) and
dedicated
communication lines in
a Community Cloud.
This offers a better
security from external
threats.
Cloud Computing
Module 86

1
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• The cloud
infrastructure is a
composition of two or
more distinct cloud
infrastructures
(private, community,
or public).
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• The hybrid cloud
components
infrastructures
(private, community,
or public) remain
unique entities.
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• The hybrid cloud
components
infrastructures
(private, community,
or public) are bound
together by
standardized or
proprietary technology
that enables data and
application portability
(for load balancing
between clouds).
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• Hybrid Clouds are
often possible when
the phenomenon of
Cloud Bursting is
applied whereby a
consumer uses a
private cloud in
routine but may use
the services of other
types of clouds for
load balancing at peak
times.
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• Hybrid Clouds are also
formed when one type
of cloud is used to
provide backup to
another type of cloud.
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• An organization may
choose to process
sensitive data on
outsourced private-
cloud but choose new
software testing on a
public cloud.
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• It may be cost
effective to put the
web requests handling
for web applications
on a PaaS instance
while the background
processing of those
web applications can
be done on on-site
community cloud.
Cloud Deployment Models: Overview of Hybrid Cloud

• Characteristics of Hybrid
Cloud according to NIST:
• Challenges for hybrid
clouds exist in security
management, identity
management and
access control of
multiple clouds etc.
• More complex
scenario arises when
the clouds are
dynamically joining
and exiting the hybrid
cloud.
Cloud Deployment Models: Overview of Hybrid Cloud

• General Characteristics
of Cloud according to
NIST:
• Network dependence
• IT skills required
• Workload locations are
hidden from consumer
• Security risks due to
end
multi-tenency
Cloud Computing
Module 87

11
Web Applications & Multitenant Technology

• Web Applications: These

are the applications
which use web
technologies (URL,
HTTP, HTML, XML) and
generally use web
browser based
interface.
• Can be modeled on the
basis of three-tier
model.
• Presentation layer
• Application layer
• Data layer
Web Applications & Multitenant Technology

• Web Application
Architecture:

Layer Implementation

Server side Client side

Presentation Web/ Web client

Application Server
Application

Data Data storage server

Web Applications & Multitenant Technology

• Web Application
Architecture:
Layer Implementation

Server side Client side

Presentation Web server Web client

Application Application server

Data Data storage server

Web Applications & Multitenant Technology

• Multi-tenant
Technology: The multi-
tenant applications
allow isolated to
simultaneous users
(tenants).
• The data and
configuration of each
user remains private to
other users.
Web Applications & Multitenant Technology

• Multi-tenant
Technology: The tenants
can customize the user
interface, business
process, data model and
access control of the
multi-tenant application.
Web Applications & Multitenant Technology

• Common Characteristics
of Multi-tenant
Applications:
• Usage isolation
• Data security
• Backup and restore is
separate for each
tenant
Web Applications & Multitenant Technology

• Common Characteristics
of Multi-tenant
Applications:
• Application upgrades
do not negatively
effect the existing
users
• Scalability in terms of
end
number of tenants
• Metered usage
• Databases, tables
and/or schema
isolation for each user
Cloud Computing
Module 88

19
Service Oriented Architecture

• (Web) Services are

independent units of
software (code) which
allow network based
machine-to-machine
interaction.
• Have no user interface.
• Process data between
the computers through
API calls.
• Examples: SOAP and
REST based web
services
Service Oriented Architecture

• Service oriented
architecture (SOA) is
usually a collection of
services (web services)
• These services
communicate with each
other for the exchange
of data and processing.
• Two or more services
may be coordinating an
activity.
Service Oriented Architecture

• Examples of web
services:
• Return the weather
conditions for a
specific zip code
• Return real-time traffic
conditions doe a road
or highway
• Return a stock price
for a particular
company
Service Oriented Architecture

• Web services are not

web pages.
• To use a web service
(which resides on a
remote server), a
program exchanges
messages with the
service.
• The user program sends
parameters (through
API call) such as zip
code to the web service
and waits for the reply.
Service Oriented Architecture

• Web services are

treated as black box by
the programmer.
• Web services are
interoperable which
means that programs
written in dissimilar
language/s than the
web-based service can
call the API functions.
Service Oriented Architecture

• Web Services: The core

technologies are:
• Web Service Description
Language (WSDL): A
markup language to
define the API of the
web service including
the functions and the
input/output messages
associated with each
function.
• Message input/output
are in the form of XML
and defined by XML
schema.
Service Oriented Architecture

• Web Services: The core

technologies are:
• The message
formatting is
according to a
common messaging
format defined by
Simple Object Access
Protocol (SOAP) or
through
Representation State
Transfer (REST).
Service Oriented Architecture

• Web Services: The core

technologies are:
• Universal Description,
Discovery and
Integration (UDDI) is a
standard which
regulates the service
registries in which
WSDL definitions can
be published so that
they can be discovered
by the users.
Service Oriented Architecture

• Cloud Service & Web

Services:
• These two are not
alike.
• Can be used
independent of each
other in a SOA.
• Cloud services are
SaaS, PaaS & IaaS
• Web services are API
Calls.
Service Oriented Architecture

• Cloud Service & Web

Services:
• Web services can be
the front door for the
cloud services running
at the backend.
• Cloud services are
often provided over
web services.
 Service Oriented Architecture

• Cloud Service & Web

Services:
• For example Amazon
Web Service (AWS)
based cloud services
(e.g., data processing
service deployed by a
provider) can be
accessed over network
end through API
developed (by the
same provider) using
Amazon API Gateway.
Cloud Computing
Module 89

31
Cloud Security Threats

• This module is about the

prominent security
threats to the Cloud
computing.
Cloud Security Threats

1. Traffic Eavesdropping:
It is possible that the
data being transferred
from Cloud consumer
to provider may get
eavesdropped
(sectary copied)
through a malicious
network-carrier.
 Cloud Security Threats

1. Traffic Eavesdropping:
Compromises the
message contents.
Illegally copied message Can go undetected for
extended periods of
time.

Cloud
Service

Cloud consumer
Traffic eavesdropper device
Cloud Security Threats

2. Malicious
Intermediately: The
messages are illegally
intercepted and then
the contents are
updated. The updated
message is then
relayed towards the
cloud.
Cloud Security Threats

2. Malicious
Intermediary: The
messages are illegally
intercepted and then
the contents are
updated. The updated
message is then
relayed towards the
cloud.
 Cloud Security Threats

2. Malicious
Intermediary: The
Original message Updated message
message may be
updated with
malicious contents
which reach the VM
hosting the cloud
service undetected.

Compromised VM

Cloud consumer
Malicious intermediary device
Cloud Computing
Module 90

38
Cloud Security Threats

3. Denial of Service
(DoS): The purpose is
to overload the IT
resources so the sage
where they can not
work properly. Can be
launched in the
following ways:
Cloud Security Threats

3. Denial of Service
(DoS):
• Workload on a cloud
service is artificially
increased through
fake messages or
repeated
communication
requests.
Cloud Security Threats

3. Denial of Service
(DoS):
• Network is overloaded
with traffic to cripple
the performance and
increasing the
response time.
Cloud Security Threats

3. Denial of Service
(DoS):
• Multiple cloud service
requests are sent.
Each request is
designed to consume
excessive memory and
processing resources.
 Cloud Security Threats

3. Denial of Service
(DoS)

Overload
Cloud service consumer A
(attacker)
VM A

VM B

Cloud service consumer B

Cloud Security Threats

4. Insufficient
Authorization based
attack: It is a situation
when a malicious user
gets direct access to IT
resources which are
supposed to be
accessed by trusted
users only.
• Happens when a
broad access is
provided to the IT
resources and/or due
to erroneously.
Cloud Security Threats

• Weak authentication
based attacks: Happen
when weak passwords
or shared (login)
accounts are used to
protect the IT
resources.
 Cloud Security Threats

The impact of attacks due

to insufficient
authorization and weak
authentication depends
upon the range of IT
resources and the range
of access to those IT
resources is
compromised.
end
Cloud Computing
Module 91

1
Cloud Security Threats

5. Virtualization Attack:
Based upon the
administrative
privileges provided to
the Cloud consumers
and multi-tenancy, it is
possible to
compromise the
underlying physical
hardware.
Cloud Security Threats

5. Virtualization Attack:
It is also possible that
the security flaws be
arising due to VM
sprawl (a lack of
security patches on OS
installed on VM).
Another possibility is
the installation of VM-
aware malware to
exploit the security
flaws of hypervisor.
Cloud Security Threats

5. Virtualization Attack:
Following are possible
sources in which the
physical server may be
compromised:
a) By an imposter in
disguise of a
legitimate consumer.
The attacker cracks
the (weak) password
of a consumer.
b) By a trusted but
malicious consumer.
Cloud Security Threats

5. Virtualization Attack:
In either case, the
vulnerabilities in the
virtualization platform
are exploited over a
single VM to take
control of the physical
server hosting the
infected VM. Makes all
the VMs hosted on the
compromised server
as vulnerable.
Cloud Security Threats

5. Virtualization Attack:
A more severe
scenario arises when
the infected VM is
migrated to other
server for load
balancing. In this case,
a number of servers
may get
compromised.
Cloud Security Threats

6. Overlapping Trust
Boundaries: Moving of
consumer data to
Cloud means that the
provider now shares
(with the consumer)
the responsibilities of
availability,
confidentiality and
integrity of data.
Cloud Security Threats

6. Overlapping Trust
Boundaries: The
consumer thus
extends the trust
boundary to include
the cloud provider.
This is prone to
vulnerabilities.
• When multiple
consumers of a cloud
share an IT resource,
the trust boundaries
overlap.
Cloud Security Threats

6. Overlapping Trust
Boundaries: The
provider may not be
able to provider the
security features that
can satisfy the security
requirement of all the
consumers of shared
IT resource on a Cloud.
 Cloud Security Threats

6. Overlapping Trust
Boundaries: More
complex scenarios
arise when the
consumer data is
replicated and stored
on multiple sites.
• Another complexity
arises when the Cloud
end provider handover the
business to a new
owner. The data
integrity becomes
threatened in both
cases.
Cloud Computing
Module 92

11
Cloud Security Threats

7. Flawed
Implementation: The
implementation of
Cloud services may
have some flaws
related to
configuration
resulting into the
occurring of
unexpected events.
Cloud Security Threats

7. Flawed
Implementation:
Particularly the
security and
operational
weaknesses in Cloud
provider’s
software/hardware
can be targeted by the
attackers to put the
integrity,
confidentiality and/or
availability of IT
resources of the
provider at stake.
Cloud Security Threats

7. Flawed
Implementation:
Equally important
point is the
implementation flaws
of Cloud services may
result in the crash of
VM and thus will
effect all the other
services on that VM as
well.
Cloud Security Threats

7. Flawed
Implementation:
For example
service A has some
implementational
flaws to crash the
hosting VM when a
A
VM certain message is
sent. This will also
B effect the services
B and C and can be
C exploited by an
attacker.
Cloud Security Threats

8. Disparity of Computer
Security Policy : A
computer security
policy defines the set
of rules and
mechanisms to ensure
the security of the
computers of the
organization. The
computer security
policies of the
consumer and
provider may not
match.
Cloud Security Threats

8. Disparity of Computer
Security Policy :
Before opting of
outsourcing and/or
public cloud, an
organization must
evaluate the
compatibility of
provider’s security
policy with its own.
Cloud Security Threats

8. Disparity of Computer
Security Policy : The
lack of administrative
privileges provided to
the consumer makes
the implementation of
the consumer chosen
computer security
policy very difficult.
 Cloud Security Threats

8. Disparity of Computer
Security Policy : Due to
the discussed points,
the standardization of
securing the IT
resources leased by a
consumer and the
consumer data is a
challenging task.
end
Cloud Computing
Module 93

20
Cloud Security Threats

9. Contracts: As an
additional
consideration, the SLA
offered by the
provider should be
carefully examined to
clarify the liabilities
taken by the provider
and the security policy
implemented by the
provider. This helps in
determining the
following:
Cloud Security Threats

9. Contracts:
• If the consumer
deploys its own
solution over the
Cloud resources then
it is a situation of
consumer’s assets
deployed over
provider’s assets.
Then how the blame
will be determined
when a security
breach or a runtime
failure occurs ?
Cloud Security Threats

9. Contracts:
• If the consumer can
apply its own security
policies while the
cloud provider keeps
the administrative
rights to the IT
infrastructure. Then
how this disparity will
be overcome.
Cloud Security Threats

10. Risk Management:

The cloud consumers
should perform a
cyclic process of risk
management to
access the potential
threats and challenges
related to Cloud
adoption. This should
be a part of risk
management strategy.
It is a three stage
process.
 Cloud Security Threats

10. Risk Management:

1. Identify the vulnerabilities of
the (to be adopted) Cloud
environment.
Identify, quantify and qualify
the risks related to how the
consumer is planning to use
the cloud.

3. Review the risks occurred

during a review period and 2. Create the security
how effective were the policies policy/ies to avoid and/or
(made in previous step). lower the impact of the risks
Identify any policy updates identified in the previous step.
required.
 Cloud Security Threats

end
Cloud Computing
Module 94

27
Trust Issues in Cloud Computing

• Link between Privacy,

Security and Trust:
• Privacy: The
confidentiality of data
related to a person or
organization.
• Security: The
preservation of
confidentiality,
integrity and
availability of data.
Trust Issues in Cloud Computing

• Link between Privacy,

Security and Trust:
• Trust: The state of
accepting a
vulnerability on the
base of positive
expectations.
Trust Issues in Cloud Computing

• Privacy issues of Cloud

Computing:
• Lack of user control
• Lack of training and
expertise
• Possibility of
secondary
(/unauthorize) use of
consumer data
• Legal compliance
Trust Issues in Cloud Computing

• Security issues of Cloud

Computing:
• Overlapping security
boundaries
• Unauthorized access
• Lack of
interoperability of
security policies
• Uncertainty of data
deletion
• Compromise of
management console
Trust Issues in Cloud Computing

• Security issues of Cloud

Computing:
• Backup vulnerabilities
• Isolation failure in
multi-tenant
applications
• Inadequate
monitoring and audit
Trust Issues in Cloud Computing

• Trust in Cloud: The

consumer’s trust in
Cloud is affected by the
privacy and security
vulnerabilities of Cloud
as discussed before.
• Further, due to lack of
transparency the
blame of
responsibility is
difficult to be placed
if the provider is
outsourcing the IT
resources from a
chain of outsourcing.
Trust Issues in Cloud Computing

• Trust in Cloud:
• The pay-as-you-go
and on-demand
provision of cloud
resources may be
subject to low level of
trust.
• The lack of trust is the
key factor for user
reluctance to use
Cloud services.
• Consumer feels a lack
of control in shifting
to Cloud.
Trust Issues in Cloud Computing

• Trust in Cloud:
• The companies
shifting from on-
premises setups to
public Clouds are
more concerned
about data security
and health than of the
servers.
• Concerns are present
regarding foreign
governments’ access
to consumers’ data
on Cloud.
Trust Issues in Cloud Computing

• Trust in Cloud:
• The analysis of
tradeoffs of Cloud
privacy, security, cost
and benefits
determines the
decision of Cloud
usage.
Trust Issues in Cloud Computing

• Trust in Cloud:
• Conclusion: The
consumers’ trust can
be assured through
the safeguarding of
personal/confidential/
sensitive data. The
existence
/enhancement of
transparency and
accountability can
increase the trust.
 Trust Issues in Cloud Computing

• Trust in Cloud:
• Conclusion: Research
should be conducted
to quantify and model
the trust and trust
management, so that
approaches for
strengthening the
consumers’ trust can
end be proposed, tested,
and/or enhanced.
Cloud Computing
Module 95

39
Cloud Infrastructure Mechanisms

• Logical Network
Perimeter: It
establishes the
boundary of virtual
network to hold with in
and isolate a set of
related cloud-IT
resources that may be
distributed physically.
• Implemented through
network devices.
Cloud Infrastructure Mechanisms

• Logical Network
Perimeter:
Implemented as virtual
environment, it has the
following components:
• Virtual Firewall to
filter the traffic of
isolated network to
and from Internet.
• Virtual Network
consisting of virtual
nodes and virtual
links.
Cloud Infrastructure Mechanisms

• Logical Network
Perimeter:

Logical Network Perimeters

VPN Service
Cloud Infrastructure Mechanisms

end
Cloud Computing
Module 96

44
Cloud Infrastructure Mechanisms

• Virtual Server: Virtual

servers or VMs emulate
the physical servers.
• Each virtual server can
host numerous IT
resources, cloud based
solutions and other
cloud computing
mechanisms.
 Cloud Infrastructure Mechanisms

• Virtual Server:
Depending upon the
Virtual servers/ capacity, a physical
Virtual Machines (VMs) server may host
multiple virtual servers.

Physical servers
 Cloud Infrastructure Mechanisms
Virtual Server/ Virtual
Machine (VM):

VM
Service
VM
SaaS consumer

IaaS consumer
Cloud Infrastructure Mechanisms

• Virtual Server: In order

to rapidly provision the
VMs with installed and
preconfigured
software such as OS,
programming
platforms etc., the
virtual servers are
cloned by templates.
Cloud Infrastructure Mechanisms

• Virtual Server: A
template is a master
copy of virtual server. It
contains the
configuration, installed
software, any
configured virtual
devices and disk
contents.
Cloud Infrastructure Mechanisms

• Virtual Server: A
consumer can:
1. Connect to a self-
service portal of Cloud
provider.
2. Choose a suitable
template.
Cloud Infrastructure Mechanisms

• Virtual Server: A
consumer can:
3. Instantiate a virtual
server through
administrative portal
which works with the
help of virtual
infrastructure
manager (VIM)
end module.
4. Customize the virtual
server through usage
and administrative
portal.
Cloud Computing
Module 97

52
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

It represents the
storage mechanisms
devised specifically for
cloud-based
provisioning.
• Instances of these
devices can be
virtualized.
• Support dynamic
scaling
• Can be accessed
remotely by Cloud
storage services.
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

The cloud storage
mechanisms support
the following (but not
limited to) logical units
of data storage:
• Files (data grouped
into files that are
located in folders)
• Blocks (the smallest
unit of data that is
individually
accessible)
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

The cloud storage
mechanisms support
the following (but not
limited to) logical units
of data storage:
• Datasets (such as
data arranged in
databases)
• Objects (data and
associated meta data)
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

Each of these levels is
associated with a
certain type of
technical interface
consisting of a specific
type of cloud storage
device with a Cloud
storage service used to
use its API.
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

1. Network Storage
Interface: For file
and block storage
2. Object Storage
Interface: Based
upon technologies
that support a range
of data and media
types. The storage
mechanism can be
accessed by REST or
SOAP based web
services.
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

3. Database Storage
Interface: Supports
the relational (SQL
based) and non-
relational databases
(NoSQL storage).
 Cloud Infrastructure Mechanisms

• Cloud Storage Device :

Database Storage Interface: Data stored in
relational database is more structured and
normalized than non-relational database. The
relational databases have higher processing
overhead. While the non-relational have high
data-redundancy. Also, transactions and joins are
not supported.
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

Database Storage Interface:
Data stored in relational
database is more structured
and normalized than non-
relational database. The
relational databases have
higher processing
overhead. While the non-
relational have high data-
redundancy. Also,
transactions and joins are
not supported.
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

Database Storage
Interface: The relational
databases have higher
processing overhead
than non-relational
database.
Cloud Infrastructure Mechanisms

• Cloud Storage Device :

Database Storage
Interface: The non-
relational databases
storage have high data-
redundancy and non-
structured data. The
relational-database
functions such as
end transactions and joins
are not supported in
non-relational database
storage.
Cloud Computing
Module 98

1
Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

It is a software used to
collect and process the
data related to Cloud-
based IT resources.
• The reporting and
analysis requirements
of the Cloud usage
module determines
the scope and volume
of data
collected/extracted.
Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

There are a few generic
types or formats of
Cloud usage monitors:
1. Monitoring Agent
2. Resource Agent
3. Polling Agent
 Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

1. Monitoring Agent: It
Log Service transparently
Database
monitors and
analyzes the
dataflow over
Monitoring Agent communication
paths. It measures
the network traffic
and messages.
 Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

2. Resource Agent:
Log VM Collects the resource
Database
usage data related to
certain events such
as initiating,
Resource Agent suspending,
resuming and
vertical scaling. It
interacts with the
Cloud resource
management
module.
 Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

Service 3. Polling Agent:
Log Collects the Cloud
Database
service usage data
after periodic polling
VM
Polling Agent
to IT resources. For
example the
uptime/downtime of
a Cloud service.
Records the updated
status of the
resource.
Cloud Computing
Module 99

7
Cloud Infrastructure Mechanisms

• Resource Replication:
It is a technique by
which multiple copies
of the IT resources are
created to increase the
availability and
productivity of the IT
resources.
Virtualization
technology is used for
Cloud IT resources’
replication.
Cloud Infrastructure Mechanisms

• Resource Replication:
For example, due to a
physical server failure
and in order to over
come the
resultant downtime of
a Cloud service
deployed over a VM
hosted by that physical
server, the entire VM
along with the
software (Cloud service
implementation) is
replicated to another
server.
Cloud Infrastructure Mechanisms

• Resource Replication:
Another example is the
horizontal scaling of IT
resources such as
increasing or
decreasing of Cloud
service instances by
replication of VM
hosting the service
instance,
corresponding to
workload.
 Cloud Infrastructure Mechanisms

• Resource Replication:
The resource
replication process
yields the IT resources
which are monitored
under the Cloud usage
monitor mechanism.
• Resource replication is
also essential for pay-
as-you-go type of
end usage & billing.
Cloud Computing
Module 100

12
Cloud Infrastructure Mechanisms

• Ready-Made
Environment: This
mechanism represents
the provisioning of
preconfigure PaaS
instances with ready to
use and customizable
programming
environments.
• Provide the
dependable PaaS
instances.
• Time efficient
provisioning
Cloud Infrastructure Mechanisms

• Ready-Made
Environment: Typically
include:
• Software
development tools
• Databases
• Middleware
• Governance tools
Cloud Infrastructure Mechanisms

• Ready-Made
Environment: The
middleware is provided
to support multi-tenant
platforms to develop
and deploy the
complementary web
services for SaaS
scenarios.
 Cloud Infrastructure Mechanisms

• Ready-Made
Environment: Overall,
the ready-made
environment
mechanism supports
the development and
production level
deployment of Cloud
services.

end
Cloud Computing
Module 101

17
Service Agreements (SAs)

• NIST identifies that the

consumer and provider
are under a legal
agreement or terms of
service.
• The agreement has two
parts:
• Service Agreement
• Service Level
Agreement (SLA)
Service Agreements (SAs)

• Service agreement
contains the legal
terms of contract.
• The SLA contains the
technical performance
promises by the
provider and the
remedies for
performance failures.
• Over all called Service
Agreements by NIST
Service Agreements (SAs)

• The following promises

are made to consumer
by the provides:
1. Availability:
• Usually 99.5% to 100%
availability is
assured.
• The assurance is for a
time intervals of a
billing cycle e.g., 15
minute, 1 hour, 1 Year
etc. for which the
service status will be
“up” for sure.
Service Agreements (SAs)

1. Availability:
• But this has to be
clarified that for
example time period of
assurance is 15 minutes
and even if the service
is “down” for 14
minutes, then it legally
means that the service
was not “down” for the
whole interval.
Service Agreements (SAs)

1. Availability:
• Typically, several
failures in subsystems
are required to
completely “down” a
service for the whole
period of billing.
• The provider may
adjust the availability
promises on case to
case basis.
Service Agreements (SAs)

2. Remedies for Failure to

Perform:
• In case of violation of
the promise of
availability (during a time
period) by the provider,
the customer will be
compensated in terms of
service credit for future
use of Cloud service.
• A refund is usually not
given.
Service Agreements (SAs)

2. Remedies for Failure to

Perform:
• Consumer is responsible
to monitor the
availability of service and
claim for compensation.
Service Agreements (SAs)

3. Remedies for Failure to

Data Preservation:
• The following situations
result in termination of
Cloud IT resources usage
for a consumer:
• Voluntarily by
consumer
• Terminated by the
provider for violating
the provider’s rule of
service and/or for non-
payment.
 Service Agreements (SAs)

3. Remedies for Failure to Data Preservation:

• The providers usually take no responsibility for
preserving the data in later case. While in former
case, the preservation is done for few days.

SAs
Consumer
Provider
Service Agreements (SAs)
4. Legal Care of Consumer
Information:
• The provider assures for
not
disclosing/viewing/using/sh
aring the consumer’s data
except in case of legal
requirement.
• On the other hand the
provider retains the right
of monitoring the
consumer data as well as
may demand a copy of
consumer’s software for
monitoring assistance.
Service Agreements (SAs)

• The following
limitations are included
in the policies by the
provider:
1. Scheduled Outages:
• Will not be
considered as service
failure.
• Will be informed in
advance.
• Will be of a limited
time period.
Service Agreements (SAs)

2. Force majeure events:

• Providers do take
the responsibility for
the events out of
their realistic
boundary. Such as:
• Power failure,
natural disaster and
unreliable
connectivity
between consumer
and cloud service.
Service Agreements (SAs)

3. Service Agreement
Changes:
• The provider usually
retain the right to
change the terms of
contract, billing
amount etc. on
limited notice.
• Consumers should
keep a regular check
for updated service
charges
Service Agreements (SAs)

3. Service Agreement
Changes:
• Sometimes the
provider inform a
specific consumer by
email or postage.
• The changes may
take effect
immediately or after
few weeks.
Service Agreements (SAs)

4. Security:
• The providers do not
take liability of data
loss, data corruption
or unauthorized data
usage if they happen
due to security breach
or due to service
interruption caused by
a malicious activity.
• At most, the service
credit is compensated
in case of data loss.
Service Agreements (SAs)

4. Security:
• Although the providers
promises for best
effort security but the
responsibility of data
security is placed on
the consumer.
• It is difficult for the
customer to determine
the cause of data loss
(malicious activity or
some other reason).
Service Agreements (SAs)

5. Service API Changes:

• The providers
generally retain the
right to delete or
update the service
API.
• Can happen any time
and without prior
notice.
Service Agreements (SAs)

• Generally the consumer

has to agree upon the
following obligations:
1. Acceptable Use
Polices: The
consumers are
generally required to
refrain from:
• Storing illegal data
• Conducting security
attacks on Cloud
infrastructure and/or
on any other user.
Service Agreements (SAs)

2. Licensed Software:
The provider require
the consumer to
install and use only the
licensed third party
software over the
Cloud.
3. Timely Payments: The
consumer should
timely pay the bill
from the provider.
Otherwise the
consumer may get
terminated after some
time.
 Service Agreements (SAs)

• Recommendations by
NIST:
• The consumers
should carefully study
and negotiate the
service agreements.
Specially take care of
the SLA assurances
and responsibilities by
the provider.
end
• Choose the most
suitable Cloud
provider periodically
after review.
Cloud Computing
Module 102

38
Cloud Hosting Data center Design

• Key terms:
• CRAC: Computer
Room Air
Conditioning
• Hot aisle
• Cold aisle
• Server cabinets
(Racks)
• Hollow floor
• Perforated tiles
Networking Structure of Cloud Hosting Data center

• Cloud hosting data

center has a layered
architecture for the
Internet access.
• The servers are
physically connected to
layer 2 switches. There
is a top of rack (TOR) in
each rack. One server is
connected to only one
TOR switch.
• The TORs are connected
to aggregate switches
(AGS).
40
Networking Structure of Cloud Hosting Data center

Internet
BR BR

AR AR

AGS AGS

TOR TOR TOR

Rack Rack Rack

41
 Cloud Hosting Data center Design

This Photo by Unknown Author is licensed under CC BY-ND

Cloud Hosting Data center Design

• Data centers consume

huge amounts of
electricity. As much as
a small town in USA.
• A large data center can
host hundreds of
thousands physical
servers.
Cloud Hosting Data center Design

• It is more costly to
setup and run a small
data center in terms of
unit costs (per server,
per MB of storage, per
GHz, Network
bandwidth) and
operational costs as
compared to larger
data centers.
Cloud Hosting Data center Design

• Google has 900,000

physical servers around
the world in its data
centers. Together
these servers consume
260 million watts of
power which accounts
to 0.01% of global
energy usage.
 Cloud Hosting Data center Design

• Facebook data center

servers process 2.4
billion pieces of
content and 750TB of
data every day.

end
Cloud Computing
Module 103

47
Data center Interconnection Networks

• The network
connecting the data
center servers is called
data center
interconnection
network.
• It is a core design of
data center.
Data center Interconnection Networks

• The network design

must support the
following features:
• Low latency
• High bandwidth
• Low cost
• Message-passing
interface (MPI)
communication
support
• Fault tolerance
Data center Interconnection Networks

• The network design

must support the
following features:
• Must satisfy both
point-to-point and
collective
communication
patterns among all
server nodes.
Data center Interconnection Networks

• Application Traffic
Support: The data
center interconnection
network must support
the MPI
communication and
high bandwidth.
• Example: Distributed
file access, Map and
Reduce functions etc.
• Some servers can be
configured to be
master and others be
slaves.
Data center Interconnection Networks

• Network Expandability:
The interconnection
network must be
expandable.
• Should support load
balancing and data
movement.
• No bottlenecks
• Can be expanded in
the unit of data center
container which
contains hundreds of
servers and is a
building block of large
data centers.
Data center Interconnection Networks

• Fault Tolerance and

Graceful Degradation:
Can be implemented
through:
• Replication in
software and
hardware resources
• Redundant links
among any two
servers
• No single point of
failure or critical links
Data center Interconnection Networks

• Fault Tolerance and

Graceful Degradation:
Can be implemented
through:
• Two layered design
should be used (a
network layer close to
servers and the upper
layer or backbone) to
support modular
end (container) based
expandable design.
Cloud Computing
Module 104

55
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
The modern data
centers are a the
collection of container
based clusters that can
be shipped from one
location to another
through trucks.
• It is an alternative to
warehouse based data
center.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• For example: The SGI
ICE Cube container can
house 46,080
processing cores or 30
PB of storage per
container.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• Such a design:
• Is more energy
efficient in terms of
cooling cost as
compared to
warehouse based
design.
• Is more mobile and
easily transportable.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• Such a design:
• Is ready to be
deployed since it is
assembled with
servers, networking,
power supplies and
cooling mechanisms.
It is then tested and
shipped.
• Helps in dynamic
scalability of data
center.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• Such a design:
• Makes the relocation
of data center as
relatively easier than
warehouse based
design.
Modular Data center and Interconnection

• Inter-Module
Connection
Networking requires an
extra layer over
modular containers to
allow dynamic scaling
and interconnection.

end
Cloud Computing
Module 105

62
Data center Management Issues

• Modern day data

centers handle ever
large volumes of data
and conduct the
processing massive
amounts of user
requests around the
globe.
Data center Management Issues

• In order to maintain
user satisfaction and
performance, the
managing of a data
center has become a
set of complex tasks.
These include (but not
limited to):
Data center Management Issues

• Making common users

happy by providing
quality services.
• Ensuring uninterrupted
and high availability of
(Cloud) services.
• Multitask
manageability:
Managing multiple
modules concurrently.
Such as processing,
networking, security
and maintenance etc.
Data center Management Issues

• Managing and planning

for the scalability of
data center.
• Ensuring the reliability
of virtual infrastructure
through fault tolerant
and recovery
mechanism to minimize
the downtime and data
loss.
 Data center Management Issues

• Managing and lowering

the operational costs
and transferring the
cost benefit to Cloud
providers and
consumers.
• Security enforcement
and data protection
• Implementation of
Green information
end
technology usage to
lower the amount of
energy consumption.
Cloud Computing
Module 98

1
Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

It is a software used to
collect and process the
data related to Cloud-
based IT resources.
• The reporting and
analysis requirements
of the Cloud usage
module determines
the scope and volume
of data
collected/extracted.
Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

There are a few generic
types or formats of
Cloud usage monitors:
1. Monitoring Agent
2. Resource Agent
3. Polling Agent
 Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

1. Monitoring Agent: It
Log Service transparently
Database
monitors and
analyzes the
dataflow over
Monitoring Agent communication
paths. It measures
the network traffic
and messages.
 Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

2. Resource Agent:
Log VM Collects the resource
Database
usage data related to
certain events such
as initiating,
Resource Agent suspending,
resuming and
vertical scaling. It
interacts with the
Cloud resource
management
module.
 Cloud Infrastructure Mechanisms

• Cloud Usage Monitor:

Service 3. Polling Agent:
Log Collects the Cloud
Database
service usage data
after periodic polling
VM
Polling Agent
to IT resources. For
example the
uptime/downtime of
a Cloud service.
Records the updated
status of the
resource.
Cloud Computing
Module 99

7
Cloud Infrastructure Mechanisms

• Resource Replication:
It is a technique by
which multiple copies
of the IT resources are
created to increase the
availability and
productivity of the IT
resources.
Virtualization
technology is used for
Cloud IT resources’
replication.
Cloud Infrastructure Mechanisms

• Resource Replication:
For example, due to a
physical server failure
and in order to over
come the
resultant downtime of
a Cloud service
deployed over a VM
hosted by that physical
server, the entire VM
along with the
software (Cloud service
implementation) is
replicated to another
server.
Cloud Infrastructure Mechanisms

• Resource Replication:
Another example is the
horizontal scaling of IT
resources such as
increasing or
decreasing of Cloud
service instances by
replication of VM
hosting the service
instance,
corresponding to
workload.
 Cloud Infrastructure Mechanisms

• Resource Replication:
The resource
replication process
yields the IT resources
which are monitored
under the Cloud usage
monitor mechanism.
• Resource replication is
also essential for pay-
as-you-go type of
end usage & billing.
Cloud Computing
Module 100

12
Cloud Infrastructure Mechanisms

• Ready-Made
Environment: This
mechanism represents
the provisioning of
preconfigure PaaS
instances with ready to
use and customizable
programming
environments.
• Provide the
dependable PaaS
instances.
• Time efficient
provisioning
Cloud Infrastructure Mechanisms

• Ready-Made
Environment: Typically
include:
• Software
development tools
• Databases
• Middleware
• Governance tools
Cloud Infrastructure Mechanisms

• Ready-Made
Environment: The
middleware is provided
to support multi-tenant
platforms to develop
and deploy the
complementary web
services for SaaS
scenarios.
 Cloud Infrastructure Mechanisms

• Ready-Made
Environment: Overall,
the ready-made
environment
mechanism supports
the development and
production level
deployment of Cloud
services.

end
Cloud Computing
Module 101

17
Service Agreements (SAs)

• NIST identifies that the

consumer and provider
are under a legal
agreement or terms of
service.
• The agreement has two
parts:
• Service Agreement
• Service Level
Agreement (SLA)
Service Agreements (SAs)

• Service agreement
contains the legal
terms of contract.
• The SLA contains the
technical performance
promises by the
provider and the
remedies for
performance failures.
• Over all called Service
Agreements by NIST
Service Agreements (SAs)

• The following promises

are made to consumer
by the provides:
1. Availability:
• Usually 99.5% to 100%
availability is
assured.
• The assurance is for a
time intervals of a
billing cycle e.g., 15
minute, 1 hour, 1 Year
etc. for which the
service status will be
“up” for sure.
Service Agreements (SAs)

1. Availability:
• But this has to be
clarified that for
example time period of
assurance is 15 minutes
and even if the service
is “down” for 14
minutes, then it legally
means that the service
was not “down” for the
whole interval.
Service Agreements (SAs)

1. Availability:
• Typically, several
failures in subsystems
are required to
completely “down” a
service for the whole
period of billing.
• The provider may
adjust the availability
promises on case to
case basis.
Service Agreements (SAs)

2. Remedies for Failure to

Perform:
• In case of violation of
the promise of
availability (during a time
period) by the provider,
the customer will be
compensated in terms of
service credit for future
use of Cloud service.
• A refund is usually not
given.
Service Agreements (SAs)

2. Remedies for Failure to

Perform:
• Consumer is responsible
to monitor the
availability of service and
claim for compensation.
Service Agreements (SAs)

3. Remedies for Failure to

Data Preservation:
• The following situations
result in termination of
Cloud IT resources usage
for a consumer:
• Voluntarily by
consumer
• Terminated by the
provider for violating
the provider’s rule of
service and/or for non-
payment.
 Service Agreements (SAs)

3. Remedies for Failure to Data Preservation:

• The providers usually take no responsibility for
preserving the data in later case. While in former
case, the preservation is done for few days.

SAs
Consumer
Provider
Service Agreements (SAs)
4. Legal Care of Consumer
Information:
• The provider assures for
not
disclosing/viewing/using/sh
aring the consumer’s data
except in case of legal
requirement.
• On the other hand the
provider retains the right
of monitoring the
consumer data as well as
may demand a copy of
consumer’s software for
monitoring assistance.
Service Agreements (SAs)

• The following
limitations are included
in the policies by the
provider:
1. Scheduled Outages:
• Will not be
considered as service
failure.
• Will be informed in
advance.
• Will be of a limited
time period.
Service Agreements (SAs)

2. Force majeure events:

• Providers do take
the responsibility for
the events out of
their realistic
boundary. Such as:
• Power failure,
natural disaster and
unreliable
connectivity
between consumer
and cloud service.
Service Agreements (SAs)

3. Service Agreement
Changes:
• The provider usually
retain the right to
change the terms of
contract, billing
amount etc. on
limited notice.
• Consumers should
keep a regular check
for updated service
charges
Service Agreements (SAs)

3. Service Agreement
Changes:
• Sometimes the
provider inform a
specific consumer by
email or postage.
• The changes may
take effect
immediately or after
few weeks.
Service Agreements (SAs)

4. Security:
• The providers do not
take liability of data
loss, data corruption
or unauthorized data
usage if they happen
due to security breach
or due to service
interruption caused by
a malicious activity.
• At most, the service
credit is compensated
in case of data loss.
Service Agreements (SAs)

4. Security:
• Although the providers
promises for best
effort security but the
responsibility of data
security is placed on
the consumer.
• It is difficult for the
customer to determine
the cause of data loss
(malicious activity or
some other reason).
Service Agreements (SAs)

5. Service API Changes:

• The providers
generally retain the
right to delete or
update the service
API.
• Can happen any time
and without prior
notice.
Service Agreements (SAs)

• Generally the consumer

has to agree upon the
following obligations:
1. Acceptable Use
Polices: The
consumers are
generally required to
refrain from:
• Storing illegal data
• Conducting security
attacks on Cloud
infrastructure and/or
on any other user.
Service Agreements (SAs)

2. Licensed Software:
The provider require
the consumer to
install and use only the
licensed third party
software over the
Cloud.
3. Timely Payments: The
consumer should
timely pay the bill
from the provider.
Otherwise the
consumer may get
terminated after some
time.
 Service Agreements (SAs)

• Recommendations by
NIST:
• The consumers
should carefully study
and negotiate the
service agreements.
Specially take care of
the SLA assurances
and responsibilities by
the provider.
end
• Choose the most
suitable Cloud
provider periodically
after review.
Cloud Computing
Module 102

38
Cloud Hosting Data center Design

• Key terms:
• CRAC: Computer
Room Air
Conditioning
• Hot aisle
• Cold aisle
• Server cabinets
(Racks)
• Hollow floor
• Perforated tiles
Networking Structure of Cloud Hosting Data center

• Cloud hosting data

center has a layered
architecture for the
Internet access.
• The servers are
physically connected to
layer 2 switches. There
is a top of rack (TOR) in
each rack. One server is
connected to only one
TOR switch.
• The TORs are connected
to aggregate switches
(AGS).
40
Networking Structure of Cloud Hosting Data center

Internet
BR BR

AR AR

AGS AGS

TOR TOR TOR

Rack Rack Rack

41
 Cloud Hosting Data center Design

This Photo by Unknown Author is licensed under CC BY-ND

Cloud Hosting Data center Design

• Data centers consume

huge amounts of
electricity. As much as
a small town in USA.
• A large data center can
host hundreds of
thousands physical
servers.
Cloud Hosting Data center Design

• It is more costly to
setup and run a small
data center in terms of
unit costs (per server,
per MB of storage, per
GHz, Network
bandwidth) and
operational costs as
compared to larger
data centers.
Cloud Hosting Data center Design

• Google has 900,000

physical servers around
the world in its data
centers. Together
these servers consume
260 million watts of
power which accounts
to 0.01% of global
energy usage.
 Cloud Hosting Data center Design

• Facebook data center

servers process 2.4
billion pieces of
content and 750TB of
data every day.

end
Cloud Computing
Module 103

47
Data center Interconnection Networks

• The network
connecting the data
center servers is called
data center
interconnection
network.
• It is a core design of
data center.
Data center Interconnection Networks

• The network design

must support the
following features:
• Low latency
• High bandwidth
• Low cost
• Message-passing
interface (MPI)
communication
support
• Fault tolerance
Data center Interconnection Networks

• The network design

must support the
following features:
• Must satisfy both
point-to-point and
collective
communication
patterns among all
server nodes.
Data center Interconnection Networks

• Application Traffic
Support: The data
center interconnection
network must support
the MPI
communication and
high bandwidth.
• Example: Distributed
file access, Map and
Reduce functions etc.
• Some servers can be
configured to be
master and others be
slaves.
Data center Interconnection Networks

• Network Expandability:
The interconnection
network must be
expandable.
• Should support load
balancing and data
movement.
• No bottlenecks
• Can be expanded in
the unit of data center
container which
contains hundreds of
servers and is a
building block of large
data centers.
Data center Interconnection Networks

• Fault Tolerance and

Graceful Degradation:
Can be implemented
through:
• Replication in
software and
hardware resources
• Redundant links
among any two
servers
• No single point of
failure or critical links
Data center Interconnection Networks

• Fault Tolerance and

Graceful Degradation:
Can be implemented
through:
• Two layered design
should be used (a
network layer close to
servers and the upper
layer or backbone) to
support modular
end (container) based
expandable design.
Cloud Computing
Module 104

55
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
The modern data
centers are a the
collection of container
based clusters that can
be shipped from one
location to another
through trucks.
• It is an alternative to
warehouse based data
center.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• For example: The SGI
ICE Cube container can
house 46,080
processing cores or 30
PB of storage per
container.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• Such a design:
• Is more energy
efficient in terms of
cooling cost as
compared to
warehouse based
design.
• Is more mobile and
easily transportable.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• Such a design:
• Is ready to be
deployed since it is
assembled with
servers, networking,
power supplies and
cooling mechanisms.
It is then tested and
shipped.
• Helps in dynamic
scalability of data
center.
Modular Data center and Interconnection

• Modular Data Center in

Shipping Containers:
• Such a design:
• Makes the relocation
of data center as
relatively easier than
warehouse based
design.
Modular Data center and Interconnection

• Inter-Module
Connection
Networking requires an
extra layer over
modular containers to
allow dynamic scaling
and interconnection.

end
Cloud Computing
Module 105

62
Data center Management Issues

• Modern day data

centers handle ever
larger volumes of data
and conduct the
processing massive
amounts of user
requests around the
globe.
Data center Management Issues

• In order to maintain
user satisfaction and
performance, the
managing of a data
center has become a
set of complex tasks.
These include (but not
limited to):
Data center Management Issues

• Making common users

happy by providing
quality services.
• Ensuring uninterrupted
and high availability of
(Cloud) services.
• Managing multiple
modules concurrently.
Such as processing,
networking, security
and maintenance etc.
Data center Management Issues

• Managing and planning

for the scalability of
data center.
• Ensuring the reliability
of virtual infrastructure
through fault tolerant
and recovery
mechanism to minimize
the downtime and data
loss.
 Data center Management Issues

• Managing and lowering

the operational costs
and transferring the
cost benefit to Cloud
providers and
consumers.
• Security enforcement
and data protection
• Implementation of
Green information
end
technology usage to
lower the amount of
energy consumption.
Cloud Computing
Module 106

1
Generic Cloud Architecture Considerations

• A generic architecture
of a (public) Cloud can
be envisioned on the
basis of technologies
we have studied so far.
• Major goals of a Cloud
platform can be:
• Scalability
• Virtualization
• Efficiency
• Reliability
Generic Cloud Architecture Considerations

• A Cloud management
software receives the
consumers’ requests
for IT resources and
provisions these
resource by using
various internal
services.
Generic Cloud Architecture Considerations

• A Cloud architecture has

to deal with certain
challenges. A few of
them are:
• Establishment of large
scale computing
(hardware + software)
infrastructure.
• User friendly and
efficient management
of Cloud
infrastructure.
Generic Cloud Architecture Considerations

• Ensuring scalability of
IT resources.
• Reliable and fault
tolerant
implementation for
processing and data.
• Implementation of
disaster recovery
mechanisms.
• Cloud architecture
should be expandable
by adding more
hardware.
Generic Cloud Architecture Considerations

• Software, hardware and

network technologies
have emerged as Cloud
enabling technologies.
Generic Cloud Architecture Considerations

• Enhancement in the
following technologies
have contributed
towards wide spread
establishment of Cloud
computing:
• Software: Virtualization,
multi-tenancy, web
applications, SOA, load
balancing, monitoring,
billing, data storage
• Hardware: CPU, memory,
storage, network
• Connectivity: Web2.0
Generic Cloud Architecture Considerations

This Photo by Unknown Author is licensed under CC BY

(https://creativecommons.org/licenses/by/3.0/)
Cloud Computing
Module 107

9
Layered Cloud Architecture

• Cloud architecture can

be considered as
consisting of layers and
sub-layers of Services
with each layer
supporting the upper
layer.
• In order of dependency,
these layers are
grouped at high level as:
• SaaS
• PaaS
• IaaS
Layered Cloud Architecture

• Software Service
development and
deployment requires a
platform service.
• A platform service is
deployed over a VM
provisioned through
IaaS.
• Some services may
draw resources from
multiple layers/sub-
layers.
Layered Cloud Architecture

• The scope of support

from vendor side is
highest for SaaS and
lowest for IaaS.
• Services developed on
PaaS require the later to
provide support for
scalability, security, and
must be dependable.
Layered Cloud Architecture

• Concur, RightNOW, Teleo,

SaaS Kenexa, Webex, Blackbaud,

salesforce.com, Netsuite,
Kenexa, etc.
• Force.com, App Engine,
Facebook, MS Azure, NetSuite,
PaaS IBM BlueCloud, SGI Cyclone,
eBay
• Amazon AWS, OpSource Cloud,

IaaS IBM Ensembles, Rackspace

cloud, Windows Azure, HP,
Banknorth
Layered Cloud Architecture

• Unless there is
interoperability among
the Clouds, a Service
deployed on a certain
platform instance may
not be portable to
another platform.
Generic Cloud Architecture Considerations

This Photo by Unknown Author is licensed under CC BY

(https://creativecommons.org/licenses/by/3.0/)
Cloud Computing
Module 108

16
Virtualization Support and Disaster Recovery

• The IT resources and

data are prone to
disasters (natural and/or
human made) which
damage them partially
or fully and thus may
crash the whole
computing system of an
organization.
Virtualization Support and Disaster Recovery

• Key terms:
• Failover: It is process
through which a system
transfers control (usually
automatedly)to an
alternate deployment
upon failure of primary
deployment.
• Failback: The process of
restoring of the system
from alternative to
primary deployment and
restoration of original
state.
Virtualization Support and Disaster Recovery

• The use of virtualization

can implement the
failover and brings
reduction in failback
time.
• As compared to (for
example) a data
disaster for data stored
on magnetic tapes, days
are require for
restoration/recovery.
Virtualization Support and Disaster Recovery

• The redundant
deployment of software
solutions, data and IT
resources is quite easy
by using virtualization.
• One deployment is
considered as primary,
while other
deployment/s are kept
as backup.
Virtualization Support and Disaster Recovery

• The backup deployment

is either updated
periodically or the
image/snapshot of the
primary deployment
(e.g., VMs) can be
saved.
• Upon failure, the
backup deployment
takes over.
• The primary
deployment is then
restored from the most
recent snapshot.
Virtualization Support and Disaster Recovery

• Virtualization has
become the core part of
disaster recovery plans
of major organizations
since last decade.
• Virtualization even
allows the testing of
disaster recovery plan
through emulation and
without disturbing the
production/primary
deployment.
Virtualization Support and Disaster Recovery

• Although the failed

physical servers have to
be re-
purchased/repaired, but
the virtualization lowers
the additional costs and
time related to failback.
• The organizations
should mark the critical
applications and data
end and use replication of
data in virtualized
environments to
support effective
disaster recovery.
Cloud Computing
Module 109

24
Cloud Architectural Design Challenges

• Challenge 1: Service
availability and Data
Lock-in Problem:
• Depending upon a
single provider for
service deployment
results in a single point
of failure or lock-in.
Cloud Architectural Design Challenges

• Challenge 1: Service
availability and Data
Lock-in Problem:
• High availability of a
service can be assured
by distributed
deployment over
multiple Clouds.
• Requires the
interoperability/standa
rdization of API calls
on different PaaS
platforms.
Cloud Architectural Design Challenges

• Challenge 2: Data
Privacy and Security
Concerns:
• Due to public access of
Clouds, multitenancy
and sophisticated
attacks/malware, the
implementation and
assurance of privacy
and security of
consumers’ data is a
big challenge.
Cloud Architectural Design Challenges

• Challenge 3:
Unpredictable
Performance and
Bottlenecks:
• The unpredictability of
processing and data
load over Cloud
services introduce I/O
bottlenecks such as
concurrent read/write
access requirements
to shared storage for
large data volumes by
multiple VMs.
Cloud Architectural Design Challenges

• Challenge 3:
Unpredictable
Performance and
Bottlenecks:
• The providers have to
carefully analyze the
deployment decisions
according to surge in
computing/data loads
and should tune the
bottlenecks.
Cloud Computing
Module 110

30
Cloud Architectural Design Challenges

• Challenge 4: Distributed
Storage and
Widespread Software
Bugs:
• Ensuring data
consistency, durability
and high availability is
a challenge when the
data is distributed.
• Debugging of data to
remove
inconsistencies and
errors is important but
challenging.
Cloud Architectural Design Challenges

• Challenge 5: Cloud
Scalability,
Interoperability and
Standardization:
• Scalability is one of the
basic features of Cloud
computing and thus
requires (for example)
dynamic availability of
IT resources
(hardware) for scaling
up.
Cloud Architectural Design Challenges

• Challenge 5: Cloud
Scalability,
Interoperability and
Standardization:
• The heterogeneity in
hardware and/or
hypervisor makes it
challenging to
dynamically include
more
hardware/virtualized
IT resources.
Cloud Architectural Design Challenges

• Challenge 5: Cloud
Scalability,
Interoperability and
Standardization:
• The open virtualization
format (OVF)
describes and open,
secure, efficient,
portable and
extensible format for
packaging and
distribution of VMs
and the software to be
deployed over VMs.
Cloud Architectural Design Challenges

• Challenge 5: Cloud
Scalability, Interoperability
and Standardization:
• OVF allows hypervisor,
guest OS and hardware
platform independent
packaging of VMs and
software.
• Interoperability should be
provided for cross
hypervisor and cross
platform (intel & AMD)
live migration of VMs.
Cloud Architectural Design Challenges

• Challenge 6: Software
Licensing and Reputation
Sharing:
• The fact that the license
model of commercial
software is not suitable
for utility computing, the
providers have to rely
upon open source
software and/or bulk
usage license.
Cloud Architectural Design Challenges

• Challenge 6: Software
Licensing and Reputation
Sharing:
• If the reputation of a
provider is affected (due
to consumers’ malicious
behavior), then there is
no service to safe-guard
the provider’s reputation.
end
Cloud Computing
Module 111

38
Public Cloud Platforms Architecture Examples

• We shall look at a few

examples of PaaS platforms
on public clouds.
 Public Cloud Platforms Architecture Examples

VM
PaaS
Service
VM
SaaS Developer

Supporting VM
Service

Supporting
Service
Public Cloud Platforms Architecture Examples

• Google App Engine (GAE):

It is a popular platform for
developing Cloud
applications.
• Based upon technologies:
• Google File System (GFS):
Stores large volumes of
data
Public Cloud Platforms Architecture Examples

• Google App Engine (GAE):

It is a popular platform for
developing Cloud
applications.
• Based upon technologies:
• MapReduce: Used in
parallel job execution on
massive data
• Chubby (Distributed
applications’ locking)
• BigTable (Storage service
to access structured data)
Public Cloud Platforms Architecture Examples

• Google App Engine (GAE):

Consumers are allowed to
develop applications in
popular languages such as
Java, PHP, Go and Python.
The following are
components of GAE:
• Datastore
• Application runtime
environment (for web
applications)
• Software Development
Kit (SDK) (for local
application development)
Public Cloud Platforms Architecture Examples

• Google App Engine (GAE):

• Administration console
(management of user
application development
cycles)
• Web service infrastructure
(interfaces for flexible use
of storage and networks
resources)
Public Cloud Platforms Architecture Examples

• Google App Engine (GAE):

• Well known applications of
GAE are Google Search
Engine, Google Docs,
Google Earth, and Gmail.
• Consumers can create
Cloud applications by using
GAE which run on Google
data centers.
Public Cloud Platforms Architecture Examples

• Amazon Web Services

(AWS):
• Amazon provides the SOAP
web services and IaaS to
the consumers/developers
to create and host Cloud
services.
• Amazon Elastic Computing
Cloud (EC2) is a web
service to provide the VMs
for hosting Cloud
applications.
Public Cloud Platforms Architecture Examples

• Amazon Web Services

(AWS):
• Simple Storage Service (S3)
provides the object-
oriented storage service.
• Elastic Block Service (EBS)
provides the block storage
interface.
• Simple Queue Service
(SQS) provides inter
process message passing.
Public Cloud Platforms Architecture Examples

• Amazon Web Services

(AWS):
• Amazon DevPay service
can be used for online
billing and account
management for the
service providers to sell
the applications developed
and/or hosted on AWS.
end
Cloud Computing
Module 112

49
Specialized Cloud Mechanisms

• Automated Scaling Listener

(ASL): It is the software
module (service agent)
which monitors and tracks
the communication
between Cloud service and
the service consumer for
dynamic scaling purpose.
• Can indicate the need for
scaling to cloud consumer.
• Indicates to cloud manager
for scaling in/out (if
configured to auto scaling
by the consumer).
 Specialized Cloud Mechanisms

H
y
Service A p
1 S e
VM L r
H v
A
y i
SaaS Consumer S
p s
Service VM L
e o
2 r
r
v
i
s
VIM o
r
 Specialized Cloud Mechanisms

H
y
A
Service p
S
1 e
VM L
r
A H v
S y i
VM L p s
Service
SaaS Consumers 2 e o
r r
v
i
s
VIM o
r
 Specialized Cloud Mechanisms

H
y
A
Service p
S
1 e
VM L
r
A H v
S y i
VM L p s
Service
SaaS Consumers e o
2
r r
v
i
s
VIM o
r
 Specialized Cloud Mechanisms

H
y
A
Service p
S
1 e
VM L
r
A H v
S y i
VM L p s
Service
SaaS Consumers e o
2
r r
v
i
s
VIM o
r
 Specialized Cloud Mechanisms

H
y
A
Service p
S
1 e
VM L
r
H v
y i
VM p s
Service
SaaS Consumers A e o
2
S r r
L v
i
s
VIM o
r
 Specialized Cloud Mechanisms

H
y
A
Service p
S
1 e
VM L
r
H v
y i
SaaS Consumers VM p s
Service
A e o
2
S r r
L v
i
s
VIM o
r
 Specialized Cloud Mechanisms

H
y
A
Service p
S
1 e
VM L
r
H v
y i
SaaS Consumers VM p s
Service
A e o
2
S r r
L v
i
s
VIM o
r
 Specialized Cloud Mechanisms

H
y
Service A p
1 S e
VM L r
H v
A
y i
SaaS Consumer S
p s
Service VM L
e o
2 r
r
v
i
s
VIM o
r
Cloud Computing
Module 113

59
Specialized Cloud Mechanisms

• Load Balancer: It is the

service agent which
distributes workload among
multiple processing
resources such as multiple
service instances.
Specialized Cloud Mechanisms

• Load Balancer:
• Workload is distributed on
the basis of:
• Processing capacity of the
IT resource
• Workload prioritization
• Content-Aware
distribution
 Specialized Cloud Mechanisms

L
Service
o A
a
d

B Replication
a
l
a
n
c Service
e A
SaaS Consumers r
Specialized Cloud Mechanisms

• SLA Monitor: Works by

pinging (for example) to a
service instance to record
the “down” status with
time.
• The statistics are used to
evaluate the extent of SLA
violation.
• Uses a polling agent
(studied before).
 Specialized Cloud Mechanisms

• Pay-per-use Monitor: It is
based upon a monitoring
agent (studied before).
• It collects the resource
usage by intercepting the
messages sent to a Cloud
service by the consumer.
• Collected data (such as
transmitted data volume,
bandwidth consumption
end
etc.) is used for billing
purpose.
Cloud Computing
Module 114

65
Specialized Cloud Mechanisms

• Failover System: This

mechanism is used to
increase the reliability and
availability of IT resources
by using redundant
implementations (for
example of Cloud services).
Specialized Cloud Mechanisms

• Failover System:
• Used for:
• Mission critical programs
• Cloud (supporting)
services which can cause
a single point of failure.
• The redundant
implementations are
actively monitored for
error detection and
unavailability of resources.
Specialized Cloud Mechanisms

• Failover System:
• Configurations:
• Active-Active
• Active-Passive
Specialized Cloud Mechanisms

• Failover System:
• Configurations:
• Active-Active: The redundant
implementation is actively
processing the workload.
• Load balancer
implementation is required.
• The failover system detects
the resource failure and
directs the load balancer to
allocate workload only to
active (redundant)
implementation.
Specialized Cloud Mechanisms

• Failover System
• Configurations:
• Active-Active: When the
failed instance is
recovered or replicated,
the failover system
directs the load balancer
to start allocating the
workload to all (including
replicated) instances.
Specialized Cloud Mechanisms

• Failover System
• Configurations:
• Active-Passive: The
redundant instance is
passive till the active
instance fails.
• The failover system
when detects a failure, it
activates a redundant
instance and redirects
the workload towards
the newly activated
instance.
 Specialized Cloud Mechanisms

• Failover System
• Configurations:
• Active-Passive:
• Upon recovery or
replication of failed
instance, the failover
system puts it to stand-
by state while the
previously activated
end instance continues to
serve as the active
instance.
Cloud Computing
Module 115

73
Specialized Cloud Mechanisms

• Failover System
• Let us see the
implementations of
Failover System.
 Failover System: Case study

Active-Active

L
Service
o A
a
d
Replication
B Failover
a System
l
a
n
c Service
e A
SaaS Consumers r
 Failover System: Case study

Active-Active

L
o
a
d
XService
A

B Failover
a System
l
a
n
c Service
e A
SaaS Consumers r
 Failover System: Case study

Active-Active

L
Service
o A
a
d

B Failover
a System
l
a
n
c Service
e A
SaaS Consumers r
 Failover System: Case study

Active-Active

L
Service
o A
a
d

B Failover
a System
l
a
n
c Service
e A
SaaS Consumers r
 Failover System: Case study

Active-Passive

Standby Instance

L
Service
o A
a
d

B Failover
a System
l
a
n
c Service
e A
SaaS Consumers r
Active Instance
 Failover System: Case study

Active-Passive

Active Instance
L
Service
o A
a
d

B Failover
a System
l
a
n

SaaS Consumers
c
e
r XService
A
 Failover System: Case study

Active-Passive

Active Instance
L
Service
o A
a
d

B Failover
a System
l
a
n
c Service
e A
SaaS Consumers r
Standby Instance
Cloud Computing
Module 116

82
Specialized Cloud Mechanisms
• Resource Cluster
Mechanism:
• The Cloud promises virtually
unlimited IT resources.
• These IT resources are
(although virtualized) but
can not be provided
through a single physical
server.
• It is obvious that the Cloud
IT resources are provisioned
from multiple physical
servers located in a single or
multiple data center/s.
Specialized Cloud Mechanisms

• Resource Cluster
Mechanism :
• The resource cluster
mechanism is used to
group multiple IT resources
so that they can be used as
a single IT resource.
• This increases the
• Computing capacity
• Load balancing
• Availability
of the clustered IT
resources.
Specialized Cloud Mechanisms

• Resource Cluster
Mechanism :
• High speed
communication links are
used to connect the
clustered IT resources for:
• Workload distribution
• Task scheduling
• Data sharing
• System synchronization
• Server clusters may or
may not have a shared
storage.
Specialized Cloud Mechanisms

• Resource Cluster
Mechanism:
• Common types:
• Server Cluster:
Consisting of physical or
virtual servers. The
virtualized clusters
support the migration of
VMs for scaling and load
balancing.
Specialized Cloud Mechanisms

• Resource Cluster
Mechanism:
• Common types:
• Database Cluster: Is used
to keep redundant
implementation of
databases. It has
features to synchronize
the data across all the
redundant instances.
• Useful for active-active
and active-passive
failover systems.
Specialized Cloud Mechanisms
• Resource Cluster
Mechanism:
• Common types:
• Large Dataset Clusters:
This type of cluster is used
to partition and distribute
large datasets without
affecting the data
integrity or computing
accuracy.
• Each node processes
workloads without any
need to
depend/communicate
with other nodes.
 Specialized Cloud Mechanisms

• Resource Cluster
Mechanism:
• Additional types:
• Load Balanced Cluster:
Implements a load
balancer mechanism
(discussed before).
• HA Cluster: Implements
a failover system
end (discussed before).
Cloud Computing
Module 117

1
Specialized Cloud Mechanisms

• Multi-Device Broker: This

mechanism is used to
transform the messages
(received from
heterogenous devices of
Cloud consumers ) into a
standard format before
conveying them to the
Cloud service.
Specialized Cloud Mechanisms

• Multi-Device Broker: The

response messages from
Cloud service are
intercepted and
transformed back to the
device specific format
before conveying to the
devices through the multi-
device broker mechanism.
 Specialized Cloud Mechanisms

Multi Service A
device
Broker

This Photo by Unknown Author is licensed This Photo by Unknown Author is licensed
under CC BY under CC BY
Specialized Cloud Mechanisms

• State Management
Database: It is a device
used to temporarily store
the state data of software
programs.
• State data can be (for
example) the configuration
and number of VMs being
employed to support a
user subscription to a PaaS
instance.
Specialized Cloud Mechanisms

• State Management
Database:
• In this way, the programs
do not use the RAM for
state-caching purposes and
thus the amount of
memory consumed is
lowered.
• The services can then be in
a “stateless” condition.
Specialized Cloud Mechanisms

• State Management
Database:
• For example, a PaaS
instance (ready-made
environment) requires
three VMs. If user pauses
activity, the state data is
saved in state management
software and the
underlying infrastructure is
scaled in to a single VM.
Specialized Cloud Mechanisms

• State Management
Database:
• When the user resumes the
activity, the state is
restored by scaling out on
the basis of data retrieved
from state management
end database.
Cloud Computing
Module 118

9
Cloud Management

• Remote Administration
System: It is a Cloud
mechanism which provides
the APIs and tools to the
providers to develop and
used online portals.
• These portals also provide
some administrative
controls to the Cloud
consumers as well.
Cloud Management

• Remote Administration
System:
• Usage and Administration
Portal:
• Management
controlling of Cloud IT
resources
• IT resources usage
reports
Cloud Management

• Remote Administration
System:
• Self-Service Portal:
• The consumer can look
at and choose various
Cloud services
• The chosen
services/package is
submitted to Cloud
provider for automated
provisioning
Cloud Management

• Remote Administration
System:
• The remote
administration console
can be used to:
• Configure and setting
cloud services
• Provision and releasing
IT resources for on-
demand usage
• Monitor cloud service
status, usage and
performance
Cloud Management

• Remote Administration
System:
• The remote administration
console can be used to:
• QoS and SLA fulfillment
monitoring
• IT-resource leasing cost
and usage fee
management
• Managing user accounts,
security credentials,
authorization and access
control
 Cloud Management

• Remote Administration
System:
• The remote
administration console
can be used to:
• Capacity planning
end • If allowed, a Cloud
consumer can create its
own front-end application
using API calls of remote
administration system.
Cloud Computing
Module 119

16
Cloud Management

• Resources Management
System: Utilizes the virtual
infrastructure manager
(VIM) for creating and
managing the virtual IT
resources.
Cloud Management

• Resources Management
System: Typical tasks
include:
• Managing the templates
used to initialize the VMs
• Allocating and releasing
the virtual IT resources
• Starting, pausing,
resuming and termination
of virtual IT resources in
response to
allocation/release of
these resources
Cloud Management

• Resources Management
System: Typical tasks
include:
• Coordination of IT
resources for resource
replication, load balancer
and failover system
• Implementation of usage
and security policies for a
Cloud service
• Monitoring the
operational conditions of
IT resources
Cloud Management

• Resources Management
System:
• These tasks can be
accessed by the cloud
resource administrators
(personnel) employed by
the cloud provider or cloud
consumer.
• The provider (and/or the
administrator staff of
provider) can access the
resource management
directly through native
VIM console.
Cloud Management

• Resources Management
System:
• The consumer (and/or
administrator staff of the
consumer) use the remote
administration
system(created by the
provider and) based upon
API calls of resource
management system.
 Cloud Management

Remote
Admin
System
Usage &
Admin
Portal
Native
R
VIM VIM
Console

VM Image
Repository
Resource Management System

This Photo by Unknown Author is This Photo by Unknown Author is

licensed under CC BY-SA licensed under CC BY-SA
Cloud Computing
Module 120

23
Cloud Management

• SLA & Billing Management

Systems:
• The SLA management
system provides features
for management and
moitoring of SLA.
• Uses a monitoring agent to
collect the SLA data on the
basis of predefined
metrics.
• The SLA monitoring agent
periodically pings the
service to evaluate the
“down” time if occurs.
Cloud Management

• SLA & Billing Management

Systems:
• The collected data is made
available to the usage and
administrative portals so
that an external and/or
internal administrator can
access the data for
querying and reporting
purposes.
• The SLA metrics monitored
are in accordance with the
SLA agreement.
Cloud Management

• SLA & Billing Management

Systems:
• The billing management
system collects and
processes the data related
to service usage.
• This data is used to
generate consumer invoice
and for accounting
purposes provider.
 Cloud Management

• SLA & Billing Management

Systems:
• The pay-as-you-go type of
billing specifically require
the usage data.
• The billing management
end
system can cater for
different pricing (pay-per-
use, flat rate, per allocation
etc.) models as well as
custom pricing models.
• Billing arrangement can be
pre-usage or post-usage.
Cloud Computing
Module 121

28
Fundamental Cloud Architectures

• Resource Pooling
Architecture: It is based
upon using one or more
resource pool in which
identical IT resources are
grouped and maintained
automatically by a system
which also ensures that the
resource pools remain
synchronized.
Fundamental Cloud Architectures

• Resource Pooling
Architecture: A few
examples of resources
pools are as follows:
1. Physical server pools
consisting of (ready to
use) networked servers
with installed OS and
other tools.
Fundamental Cloud Architectures

• Resource Pooling
Architecture: A few
examples of resources
pools are as follows:
2. VM (virtual server) pool/s
configured by using one
or more templates
selected by the consumer
during provisioning.
Fundamental Cloud Architectures

• Resource Pooling
Architecture: A few
examples of resources
pools are as follows:
3. Cloud storage pools
consisting of file/block
based storage structures.
4. Network pools consist of
different (preconfigured)
network connecting
devices that are created
for redundant
connectivity, load
balancing and link
aggregation.
Fundamental Cloud Architectures

• Resource Pooling
Architecture: A few
examples of resources
pools are as follows:
5. CPU pools are ready to be
allocated to VMs by the
multiple of single core.
• Dedicated pools can be
created for each type of IT
resources.
• Individual resource pools
can become sub-groups
into larger pool.
Fundamental Cloud Architectures

• Resource Pooling
Architecture:
• A resource pool can be
divided into sibling
pools as well as nested
pools.
• Sibling pools are
independent and
isolated from each
other. May have
different types of IT
resources.
Fundamental Cloud Architectures

• Resource Pooling
Architecture:
• Nested pools are
drawn from a bigger
pool and consist of the
same types of IT
resources as are
present in the parent
pool.
Fundamental Cloud Architectures
• Resource Pooling
Architecture:
• Resource pools created
for different consumers
are isolated from each
other.
• The additional
mechanisms associated
with resource pooling
are:
1. Audit monitor: Tracks
the credentials of
consumers when they
login for IT resource
usage.
Fundamental Cloud Architectures

• Resource Pooling
Architecture:
2. Cloud Usage Monitor
3. Hypervisor
4. Logical Network
Perimeter
end 5. Pay-Per-Use Monitor
6. Remote
Administration System
7. Resource
Management System
8. Resource Replication
Cloud Computing
Module 122

38
Fundamental Cloud Architectures

• Dynamic Scalability
Architecture: