Distributed Computing

Lesson 2: Distributed Systems

Thomas Weise · 汤卫思
tweise@[Link] · [Link]
Hefei University, South Campus 2 合肥学院 南艳湖校区/南2区
Faculty of Computer Science and Technology 计算机科学与技术系
Institute of Applied Optimization 应用优化研究所
230601 Shushan District, Hefei, Anhui, China 中国 安徽省 合肥市 蜀山区 230601
Econ. & Tech. Devel. Zone, Jinxiu Dadao 99 经济技术开发区 锦绣大道99号
Outline

1 Distributed Systems

2 General Features

3 Pros and Cons

4 Design Principles

website
Distributed Computing Thomas Weise 2/28
Introduction

• We will get an understanding of what distributed systems are

Distributed Computing Thomas Weise 3/28

Introduction

• We will get an understanding of what distributed systems are

• We will get an understanding of opportunities and challenges in
distributed systems

Distributed Computing Thomas Weise 3/28

Introduction

• We will get an understanding of what distributed systems are

• We will get an understanding of opportunities and challenges in
distributed systems
• We will learn some of the basic design principles for distributed
systems

Distributed Computing Thomas Weise 3/28

Distributed Systems

What is a distributed system?

What are distributed algorithms?

Distributed Computing Thomas Weise 4/28

Distributed Systems

• Various slightly different definitions exist [1–5] , but let us settle for the
following:

Distributed Computing Thomas Weise 4/28

Distributed Systems

• Various slightly different definitions exist [1–5] , but let us settle for the
following:
• A distributed system is a set of autonomous systems (nodes,
computers) which are connected by a network and communicate via
the exchange of messages.

Distributed Computing Thomas Weise 4/28

Distributed Systems

• Various slightly different definitions exist [1–5] , but let us settle for the
following:
• A distributed system is a set of autonomous systems (nodes,
computers) which are connected by a network and communicate via
the exchange of messages.
• Distributed algorithms are algorithms which can be executed by
multiple computers in a distributed system and cooperatively try to
solve a given problem.

Distributed Computing Thomas Weise 4/28

Programming of Distributed Systems

• In courses on computer programming, you have learned that it is

impossible to avoid programming mistakes.

Distributed Computing Thomas Weise 5/28

Programming of Distributed Systems

• In courses on computer programming, you have learned that it is

impossible to avoid programming mistakes.
• You have learned how important it is to test and debug programs.

Distributed Computing Thomas Weise 5/28

Programming of Distributed Systems

• In courses on computer programming, you have learned that it is

impossible to avoid programming mistakes.
• You have learned how important it is to test and debug programs.
• This is also true for programming distributed applications.

Distributed Computing Thomas Weise 5/28

Programming of Distributed Systems

• In courses on computer programming, you have learned that it is

impossible to avoid programming mistakes.
• You have learned how important it is to test and debug programs.
• This is also true for programming distributed applications.
• With the exception that there now are several entirely new possible
sources of errors.

Distributed Computing Thomas Weise 5/28

Programming of Distributed Systems

• In courses on computer programming, you have learned that it is

impossible to avoid programming mistakes.
• You have learned how important it is to test and debug programs.
• This is also true for programming distributed applications.
• With the exception that there now are several entirely new possible
sources of errors.
• Yes, this is going to get scary, be prepared.

Distributed Computing Thomas Weise 5/28

Network of Computers

• Multiple independent computers connected by communication

network

Distributed Computing Thomas Weise 6/28

Computers/Links Fail

• Computers and communication links may fail independently from

each other (and without obvious reason)

Distributed Computing Thomas Weise 7/28

Latency and Unreliability

• Information exchange only via message exchange

Distributed Computing Thomas Weise 8/28

Latency and Unreliability

• Information exchange only via message exchange

Distributed Computing Thomas Weise 8/28

Latency and Unreliability

• Information exchange only via message exchange

• Message latencies not deterministic, messages may take over each
other (definite physical limit: speed of light)

Distributed Computing Thomas Weise 8/28

Latency and Unreliability

• Information exchange only via message exchange

• Message latencies not deterministic, messages may take over each
other (definite physical limit: speed of light)
• Network may be unreliable, messages may get lost/modified

Distributed Computing Thomas Weise 8/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS

Distributed Computing Thomas Weise 9/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS:

1 Multiple processes can never be observed simultaneously

Distributed Computing Thomas Weise 9/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS:

1 Multiple processes can never be observed simultaneously

Distributed Computing Thomas Weise 9/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS:

1 Multiple processes can never be observed simultaneously

Distributed Computing Thomas Weise 9/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS:

1 Multiple processes can never be observed simultaneously

Distributed Computing Thomas Weise 9/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS:

1 Multiple processes can never be observed simultaneously
2 It is difficult to make statements about the global system state

Distributed Computing Thomas Weise 9/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS:

1 Multiple processes can never be observed simultaneously
2 It is difficult to make statements about the global system state

Distributed Computing Thomas Weise 9/28

Uncertainty: No Global View

• Friedemann Mattern [3] : Uncertainty Principles of DS:

1 Multiple processes can never be observed simultaneously
2 It is difficult to make statements about the global system state
3 Nodes have no global information and act only based on local
information → dangerous.

Distributed Computing Thomas Weise 9/28

No Global Time

• From this, it follows that: We have no common, global time

Distributed Computing Thomas Weise 10/28

No Global Time

• From this, it follows that: We have no common, global time

• Synchronization of clocks complicated because of message latency

Distributed Computing Thomas Weise 10/28

No Global Time

• From this, it follows that: We have no common, global time

• Synchronization of clocks complicated because of message latency

Distributed Computing Thomas Weise 10/28

No Global Time

• From this, it follows that: We have no common, global time

• Synchronization of clocks complicated because of message latency

Distributed Computing Thomas Weise 10/28

No Global Time

• From this, it follows that: We have no common, global time

• Synchronization of clocks complicated because of message latency

Distributed Computing Thomas Weise 10/28

No Global Time

• From this, it follows that: We have no common, global time

• Synchronization of clocks complicated because of message latency

What can happen here?

Distributed Computing Thomas Weise 10/28

No Global Time

• From this, it follows that: We have no common, global time

• Synchronization of clocks complicated because of message latency

The compiler does not see that

the file has been updated and
thus does not recompile the
project.

Distributed Computing Thomas Weise 10/28

General Features

• No globally shared memory

Distributed Computing Thomas Weise 11/28

General Features

• No globally shared memory

• no node knows the global state

Distributed Computing Thomas Weise 11/28

General Features

• No globally shared memory

• no node knows the global state
• shared memory may be emulated, though that’s not easy

Distributed Computing Thomas Weise 11/28

General Features

• No globally shared memory

• no node knows the global state
• shared memory may be emulated, though that’s not easy
• No accurate failure detection

Distributed Computing Thomas Weise 11/28

General Features

• No globally shared memory

• no node knows the global state
• shared memory may be emulated, though that’s not easy
• No accurate failure detection
• in asynchronous distributed systems without upper bound for
communication delay, it is impossible to distinguish failed from slow
processes[6]

Distributed Computing Thomas Weise 11/28

Pros and Cons

Ok, that was scary.

Distributed Computing Thomas Weise 12/28

Pros and Cons

Ok, that was scary.

Surely, these are the only bad things that I have to consider when building
a distributed application, right?

Distributed Computing Thomas Weise 12/28

Pros and Cons

• Distributed systems are complex

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems: true parallelism

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems: true parallelism,
indeterminism

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems: true parallelism,
indeterminism
• Programming them is more complex, too

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems: true parallelism,
indeterminism
• Programming them is more complex, too:
• You have to deal with parallelism.

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems: true parallelism,
indeterminism
• Programming them is more complex, too:
• You have to deal with parallelism.
• You have to deal with communication.

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems: true parallelism,
indeterminism
• Programming them is more complex, too:
• You have to deal with parallelism.
• You have to deal with communication.
• How to do testing and debugging?

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Distributed systems are complex:

• heterogeneity of hardware/OSes/applications: documentation
absolutely vital! [7]
• consist of separate (autonomous) systems: true parallelism,
indeterminism
• Programming them is more complex, too:
• You have to deal with parallelism.
• You have to deal with communication.
• How to do testing and debugging?
• Failures are often not reproducible.

Distributed Computing Thomas Weise 13/28

Pros and Cons

• Security in distributed systems is complex

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]
• Systems may span over country borders, different administrative
domains

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]
• Systems may span over country borders, different administrative
domains:
• data security

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]
• Systems may span over country borders, different administrative
domains:
• data security
• legislative security

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]
• Systems may span over country borders, different administrative
domains:
• data security
• legislative security
• central administration impossible

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]
• Systems may span over country borders, different administrative
domains:
• data security
• legislative security
• central administration impossible
• Distributed systems are expensive

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]
• Systems may span over country borders, different administrative
domains:
• data security
• legislative security
• central administration impossible
• Distributed systems are expensive:
• network infrastructure

Distributed Computing Thomas Weise 14/28

Pros and Cons

• Security in distributed systems is complex:

• confidentiality
• more attack points, DDOS attacks, intrusion detection, . . .
• even professional organizations whose business it is to be
secure. . . make terrible mistakes! [8–10]
• Systems may span over country borders, different administrative
domains:
• data security
• legislative security
• central administration impossible
• Distributed systems are expensive:
• network infrastructure
• energy consumption

Distributed Computing Thomas Weise 14/28

Leslie Lamport’s Statement

Ok, let’s see what Leslie Lamport — one of the big guys in distributed
systems – has to say.

Distributed Computing Thomas Weise 15/28

Leslie Lamport’s Statement

Ok, let’s see what Leslie Lamport — one of the big guys in distributed
systems – has to say. He has done lots of useful research [11–13] .

Distributed Computing Thomas Weise 15/28

Leslie Lamport’s Statement

Ok, let’s see what Leslie Lamport — one of the big guys in distributed
systems – has to say. He has done lots of useful research [11–13] (and also
invented LATEX [15] ).

Distributed Computing Thomas Weise 15/28

Leslie Lamport’s Statement

Ok, let’s see what Leslie Lamport — one of the big guys in distributed
systems – has to say. He has done lots of useful research [11–13] (and also
invented LATEX [15] , which sits on top of the TEX [16] system created by
Knuth).

Distributed Computing Thomas Weise 15/28

Leslie Lamport’s Statement

Ok, let’s see what Leslie Lamport — one of the big guys in distributed
systems – has to say. He has done lots of useful research [11–13] (and also
invented LATEX [15] , which sits on top of the TEX [16] system created by
Knuth, who is the guy who brought you The Art of Computer
Programming [17–19] and Concrete Mathematics [20] ).

Distributed Computing Thomas Weise 15/28

Leslie Lamport’s Statement

Ok, let’s see what Leslie Lamport — one of the big guys in distributed
systems – has to say. He has done lots of useful research [11–13] (and also
invented LATEX [15] , which sits on top of the TEX [16] system created by
Knuth, who is the guy who brought you The Art of Computer
Programming [17–19] and Concrete Mathematics [20] ).
. . . Anyway, what does he have to say? [14]

Distributed Computing Thomas Weise 15/28

Leslie Lamport’s Statement

Received: by [Link] (5.54.3/4.7.34) id AA09105; Thu, 28 May 87 [Link] PDT

Date: Thu, 28 May 87 [Link] PDT
From: lamport (Leslie Lamport) [14] To: src-t
Message-Id: <8705281923.AA09105@[Link]>
Subject: distribution
There has been considerable debate over the years about what
constitutes a distributed system. It would appear that the following
definition has been adopted at SRC:
A distributed system is one in which the failure of a computer
you didn’t even know existed can render your own computer
unusable.
The current electrical problem in the machine room is not the
culprit--it just highlights a situation that has been getting
progressively worse. It seems that each new version of the nub makes
my FF more dependent upon programs that run elsewhere.
Having to wait a few seconds for a program to be swapped in is a lot
less annoying than having to wait an hour or two for someone to reboot
the servers. I therefore propose a development project to make our
system more robust. I am not proposing any particular approach
(enabling stand-alone operation is just one possibility).
I will begin the effort by volunteering to gather some data on the
problem. If you know of any instance of user’s FF becoming inoperative
through no fault of its own, please send me a message indicating the
user, the time, and the cause (if known).
Leslie

Distributed Computing Thomas Weise 15/28

Pros and Cons

• Leslie Lamport:
“A distributed system is one in which the failure of a computer you
didn’t even know existed can render your own computer unusable.” [14]

Distributed Computing Thomas Weise 16/28

Pros and Cons

• Leslie Lamport:
“A distributed system is one in which the failure of a computer you
didn’t even know existed can render your own computer unusable.” [14]
• Verı́ssimo and Rodrigues:
“If you do not need a distributed system, do not distribute.” [5]

Distributed Computing Thomas Weise 16/28

Pros and Cons

• Leslie Lamport:
“A distributed system is one in which the failure of a computer you
didn’t even know existed can render your own computer unusable.” [14]
• Verı́ssimo and Rodrigues:
“If you do not need a distributed system, do not distribute.” [5]

So why do we even bother?

Distributed Computing Thomas Weise 16/28

Pros and Cons

• Leslie Lamport:
“A distributed system is one in which the failure of a computer you
didn’t even know existed can render your own computer unusable.” [14]
• Verı́ssimo and Rodrigues:
“If you do not need a distributed system, do not distribute.” [5]

So why do we even bother?

Are there advantages??

Distributed Computing Thomas Weise 16/28

Pros and Cons

• Leslie Lamport:
“A distributed system is one in which the failure of a computer you
didn’t even know existed can render your own computer unusable.” [14]
• Verı́ssimo and Rodrigues:
“If you do not need a distributed system, do not distribute.” [5]

So why do we even bother?

Are there advantages??

Why am I taking this course?

Distributed Computing Thomas Weise 16/28

Pros and Cons

• Some services need to carried out at a specific location

Distributed Computing Thomas Weise 17/28

Pros and Cons
• Some services need to carried out at a specific location, e.g.,
• Online ordering/manufacturing systems

Distributed Computing Thomas Weise 17/28

Pros and Cons
• Some services need to carried out at a specific location, e.g.,
• Online ordering/manufacturing systems
• Control of hard-to-reach systems (chemical plant, robotic moon
mission)

Distributed Computing Thomas Weise 17/28

Pros and Cons
• Some services need to carried out at a specific location, e.g.,
• Online ordering/manufacturing systems
• Control of hard-to-reach systems (chemical plant, robotic moon
mission)
• Drones in modern warfare

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature, e.g.,
• Team of robots in robotic soccer or robots in a warehous

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature, e.g.,
• Team of robots in robotic soccer or robots in a warehous
• Storage of photos/music/videos and access from multiple different
devices

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature, e.g.,
• Team of robots in robotic soccer or robots in a warehous
• Storage of photos/music/videos and access from multiple different
devices
• Access of websites/information created at different locations

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility, e.g.,
• Services in a SOA are single modules which can be maintained and
replaced separately

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility, e.g.,
• Services in a SOA are single modules which can be maintained and
replaced separately
• Computers can easily be added to clusters or grids

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility, e.g.,
• Services in a SOA are single modules which can be maintained and
replaced separately
• Computers can easily be added to clusters or grids
• Elastic cloud services can be replicated to more computers/virtual
machines added to applications when more computational power
becomes necessary

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability, e.g.,
• New components can be added to a SOA

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability, e.g.,
• New components can be added to a SOA
• Heterogeneous systems: legacy applications + new applications

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability, e.g.,
• New components can be added to a SOA
• Heterogeneous systems: legacy applications + new applications
• Purchase of more computing power in a cloud

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability, e.g.,
• New components can be added to a SOA
• Heterogeneous systems: legacy applications + new applications
• Purchase of more computing power in a cloud
• Adding of sensors to an existing manufacturing surveillance plant
system

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing, e.g.,
• Cloud and cluster computing

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing, e.g.,
• Cloud and cluster computing
• Seti@home [21] , BOINC [22]

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing, e.g.,
• Replication of web servers

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing, e.g.,
• Replication of web servers
• Access node of cluster decides where the job should be done

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility, e.g.,
• Replication

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility, e.g.,
• Replication
• Maintenance centralized

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility, e.g.,
• Replication
• Maintenance centralized
• Internet accessible from hand helds

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility, e.g.,
• Replication
• Maintenance centralized
• Internet accessible from hand helds
• Price checks from your mobile phone

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . ), e.g.,
• Large-scale cluster with central cooling and energy supply

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . ), e.g.,
• Large-scale cluster with central cooling and energy supply
• Shared infrastructure

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . ), e.g.,
• Large-scale cluster with central cooling and energy supply
• Shared infrastructure
• Computers turned on/off on demand

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing, e.g.,
• US/Europe outsource many IT jobs to India, because it is cheap

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing, e.g.,
• US/Europe outsource many IT jobs to India, because it is cheap
• US + China could work in shifts, everyone works at day time, but work
is done 24h

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing, e.g.,
• US/Europe outsource many IT jobs to India, because it is cheap
• US + China could work in shifts, everyone works at day time, but work
is done 24h
• Infrastructure in US is rarely used at night → which is daytime in India

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing
• Fault tolerance / safety / replication / component replacement

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing
• Fault tolerance / safety / replication / component replacement, e.g.,
• Large companies replicate their databases (e.g., with engineering
designs) at different geographical locations

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing
• Fault tolerance / safety / replication / component replacement, e.g.,
• Large companies replicate their databases (e.g., with engineering
designs) at different geographical locations
• Critical computing infrastructure should be replicated

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing
• Fault tolerance / safety / replication / component replacement
• Decentralization

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing
• Fault tolerance / safety / replication / component replacement
• Decentralization, e.g.,
• Content “closer” to destination (see Content Delivery Networks)

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing
• Fault tolerance / safety / replication / component replacement
• Decentralization, e.g.,
• Content “closer” to destination (see Content Delivery Networks)
• Services can be carried out in different locations: less computational
load on central hardware

Distributed Computing Thomas Weise 17/28

Pros and Cons

• Some services need to carried out at a specific location

• Some tasks are distributed by nature
• Modularity and flexibility
• Incremental growth / scalability
• Resource sharing
• Load sharing / balancing / grid computing
• High availability, mobile accessibility
• Lower costs ($/MHz, $/(IO/s), . . . )
• Outsourcing / time zone advantages / cloud computing
• Fault tolerance / safety / replication / component replacement
• Decentralization
• Distribution over different jurisdictional and/or political areas

Distributed Computing Thomas Weise 17/28

Design Principles

• Only distribute if there is a striking benefit, otherwise use centralized

solutions.

Distributed Computing Thomas Weise 18/28

Design Principles

• Only distribute if there is a striking benefit, otherwise use centralized

solutions.
• What else?

Distributed Computing Thomas Weise 18/28

Single Point of Failure?

• One central server dealing with all requests and forwarding all
messages between the nodes. Is this good or bad?

Distributed Computing Thomas Weise 19/28

Single Point of Failure?

• Bad! You should avoid central/single points of failure!

Distributed Computing Thomas Weise 19/28

Single Point of Failure?

• Bad! You should avoid central/single points of failure!

Distributed Computing Thomas Weise 19/28

Single Point of Failure?

• Bad! You should avoid central/single points of failure!

• regardless which node or connection fails, the system should remain
intact

Distributed Computing Thomas Weise 19/28

Single Point of Failure?

• Bad! You should avoid central/single points of failure!

• regardless which node or connection fails, the system should remain
intact
• node failure/churn should lead to gentle performance degeneration
rather than total failure

Distributed Computing Thomas Weise 19/28

Single Point of Failure?

• Bad! You should avoid central/single points of failure!

• regardless which node or connection fails, the system should remain
intact
• node failure/churn should lead to gentle performance degeneration
rather than total failure
• bottlenecks should be avoided

Distributed Computing Thomas Weise 19/28

Scalability

• How should a system behave when the numbers of users or processors

increase?

Distributed Computing Thomas Weise 20/28

Scalability

• Scalability

Distributed Computing Thomas Weise 20/28

Scalability

• Scalability
• growth of number of tasks / users / work load / available nodes should
be anticipated

Distributed Computing Thomas Weise 20/28

Scalability

• Scalability
• growth of number of tasks / users / work load / available nodes should
be anticipated
• system performance should rise in case of more computing power

Distributed Computing Thomas Weise 20/28

Scalability

• Scalability
• growth of number of tasks / users / work load / available nodes should
be anticipated
• system performance should rise in case of more computing power
• system performance should degenerate gently in case of higher work
load

Distributed Computing Thomas Weise 20/28

Security

• Security: Information hiding / need to know

Distributed Computing Thomas Weise 21/28

Security

• Security: Information hiding / need to know

• only the information absolutely need to solve a task is handed to the
notes

Distributed Computing Thomas Weise 21/28

Security

• Security: Information hiding / need to know

• only the information absolutely need to solve a task is handed to the
notes
• security cannot be “integrated later”, security is no side dish

Distributed Computing Thomas Weise 21/28

Security

• Security: Information hiding / need to know

• only the information absolutely need to solve a task is handed to the
notes
• security cannot be “integrated later”, security is no side dish
• security by obfuscation does not work, use well-known algorithms and
methods instead

Distributed Computing Thomas Weise 21/28

Security

• Security: Information hiding / need to know

• only the information absolutely need to solve a task is handed to the
notes
• security cannot be “integrated later”, security is no side dish
• security by obfuscation does not work, use well-known algorithms and
methods instead
• Maybe you should also take an information
systems security course as well. . .

Distributed Computing Thomas Weise 21/28

Transparency

• Transparency of an aspect = the aspect is not visible to the app/user

Distributed Computing Thomas Weise 22/28

Transparency

• Transparency of an aspect = the aspect is not visible to the app/user

• hide complexity (for instance: protocol stack, middleware)

Distributed Computing Thomas Weise 22/28

Transparency

• Transparency of an aspect = the aspect is not visible to the app/user

• hide complexity (for instance: protocol stack, middleware)
• many applications of transparency in distributed systems (distribution,
location, machine, fault, replication, migration, . . . )

Distributed Computing Thomas Weise 22/28

Transparency

• Transparency of an aspect = the aspect is not visible to the app/user

• hide complexity (for instance: protocol stack, middleware)
• many applications of transparency in distributed systems (distribution,
location, machine, fault, replication, migration, . . . )
• sometimes not appropriate (system management, context aware
systems, adaptive systems, . . . )

Distributed Computing Thomas Weise 22/28

Summary

• Distributed System = autonomous nodes communicating via

messages
• Several advantages and drawbacks: distributed computing only if
necessary
• Several design principles to consider
• Scalability is always limited
• Message complexity should be low

Distributed Computing Thomas Weise 23/28

 谢谢
Thank you
Thomas Weise [汤卫思]
tweise@[Link]
[Link]

Hefei University, South Campus 2

Institute of Applied Optimization
Shushan District, Hefei, Anhui,
China
Caspar David Friedrich, “Der Wanderer über dem Nebelmeer”, 1818
[Link]

Distributed Computing Thomas Weise 24/28

Bibliography

Distributed Computing Thomas Weise 25/28

Bibliography I
1. Andrew Stuart Tanenbaum. Distributed Operating Systems. Prentice-Hall International Editions. Upper Saddle River, NJ,
USA: Prentice Hall International Inc., 1995. ISBN 0-13-143934-0, 0-13-219908-4, 978-0-13-143934-4, and
978-0-13-219908-7. URL [Link]
2. George F. Coulouris, Jean Dollimore, and Tim Kindberg. Distributed Systems: Concepts and Design. Upper Saddle River,
NJ, USA: Pearson Education and Boston, MA, USA: Addison-Wesley Longman Publishing Co., Inc., 4th rev. edition, June
2005. ISBN 0201180596, 0321263545, 9780201180596, and 9780321263544. URL
[Link]
3. Friedemann Mattern. Verteilte Basisalgorithmen, volume 226 of Informatik-Fachberichte (IFB). Berlin, Germany:
Springer-Verlag GmbH, October 1989. ISBN 0-387-51835-5, 3540518355, 978-0-387-51835-0, and 9783540518358. URL
[Link] Based on his dissertation at Prof. Dr. J. Nehmer’s group /
Sonderforschungsbereich “VLSI-Entwurf und Parallelität” at the computer science department of the University of
Kaiserslautern.
4. Andrew Stuart Tanenbaum and Maarten van Steen. Distributed Systems. Principles and Paradigms. Upper Saddle River,
NJ, USA: Prentice Hall International Inc. and Upper Saddle River, NJ, USA: Pearson Education, March 2003. ISBN
0130888931, 0131217860, 0132392275, 8178087898, 9780130888938, 978-0131217867, 9780132392273, and
9788178087894. URL [Link]
5. Paulo Verı́ssimo and Luı́s Rodrigues. Distributed Systems for System Architects, volume 1 of Advances in Distributed
Computing and Middleware. Berlin, Germany: Springer-Verlag GmbH, 2001. ISBN 0792372662 and 9780792372660. URL
[Link]
6. Michael J. Fischer, Nancy A. Lynch, and Michael S. Paterson. Impossibility of distributed consensus with one faulty
process. Journal of the Association for Computing Machinery (JACM), 32(2):374–382, April 1985. doi:
10.1145/3149.214121. URL [Link]
7. Barry M. Leiner, Vinton G. Cerf, David D. Clark, Robert E. Kahn, Leonard Kleinrock, Daniel C. Lynch, Jonathan B. Postel,
Lawrence G. Roberts, and Stephen Wolff. Brief history of the internet. ACM SIGCOMM Computer Communication
Review, 39(5):22–31, October 2009. doi: 10.1145/1629607.1629613. URL
[Link]
8. Adam Langley and David G. Andersen. [Link]. Mountain View, CA, USA: Google Inc. URL
[Link]

Distributed Computing Thomas Weise 26/28

Bibliography II
9. Adam Goodman. Bypassing google’s two-factor authentication. The Duo Bulletin: Notes on security, from the desks of
Duo Security, February 25, 2013. URL
[Link]
10. Michael Messner. Multiple vulnerabilities in d’link dir-600 and dir-300 (rev b). s3cur1ty: no one is safe. . . , February 4, 2013.
11. Leslie Lamport. Time, clocks, and the ordering of events in a distributed system. Communications of the ACM (CACM),
21(7):558–565, July 1978. doi: 10.1145/359545.359563. URL
[Link] Also: Report CA-7603-2911, Massachusetts
Comptuter Association, Wakefield, Massachusetts, USA, March 1976.
12. Leslie Lamport, Robert Shostak, and Marshall Pease. The byzantine generals problem. ACM Transactions on Programming
Languages and Systems (TOPLAS), 4(3):382–401, July 1982. doi: 10.1145/357172.357176. URL
[Link]
13. Leslie Lamport. A new solution of dijkstra’s concurrent programming problem. Communications of the ACM (CACM), 17
(8):453–455, August 1974. doi: 10.1145/361082.361093. URL
[Link]
14. Leslie Lamport. Subject: distribution, May 28, 1987. URL
[Link]
15. Leslie Lamport. LaTeX: A Document Preparation System. User’s Guide and Reference Manual. Reading, MA, USA:
Addison-Wesley Publishing Co. Inc., 1994. ISBN 0201529831 and 9780201529838. URL
[Link]
16. Donald Ervin Knuth. The TeXbook (Computers and Typesetting, Volume A). Reading, MA, USA: Addison-Wesley
Publishing Co. Inc., 1984. ISBN 0-201-13448-9. URL
[Link]
17. Donald Ervin Knuth. Sorting and Searching, volume 3 of The Art of Computer Programming (TAOCP). Boston, MA, USA:
Addison-Wesley Longman Publishing Co., Inc., 2nd edition, 1998. ISBN 0-201-03803-X, 0-201-03809-9, 0-201-03822-6,
0-201-85394-9, 0-201-89685-0, 978-0-201-03803-3, 978-0-201-03809-5, 978-0-201-03822-4, 978-0-201-85394-0, and
978-0-201-89685-5. URL [Link] Original from the University of Michigan.
18. Donald Ervin Knuth. Fundamental Algorithms, volume 1 of The Art of Computer Programming (TAOCP). Boston, MA,
USA: Addison-Wesley Longman Publishing Co., Inc., third edition, 1997. ISBN 0-201-89683-4 and 978-0-201-89683-1.
URL [Link]

Distributed Computing Thomas Weise 27/28

Bibliography III
19. Donald Ervin Knuth. Seminumerical Algorithms, volume 2 of The Art of Computer Programming (TAOCP). Boston, MA,
USA: Addison-Wesley Longman Publishing Co., Inc., 1969. ISBN 0-201-89684-2, 8177583352, 978-0-201-89684-8, and
978-8177583359. URL [Link]
20. Ronald Graham, Donald Ervin Knuth, and Oren Patashnik. Concrete Mathematics: A Foundation for Computer Science.
Reading, MA, USA: Addison-Wesley Publishing Co. Inc., 1988. ISBN 0-201-14236-8 and 0-201-55802-5.
21. Eric Korpela, Dan Werthimer, David Anderson, Jeff Cobb, and Matt Lebofsky. Seti@home – massively distributed
computing for seti. Computing in Science & Engineering, 3(1):78–83, January–February 2001. doi:
10.1109/5992.895191. URL [Link]
22. Miguel A. Vega-Rodrı́guez, David Vega-Pérez, Juan A. Gómez-Pulido, and Juan M. Sánchez-Pérez. Radio network design
using population-based incremental learning and grid computing with boinc. In Mario Giacobini, Anthony Brabazon,
Stefano Cagnoni, Gianni A. Di Caro, Rolf Drechsler, Muddassar Farooq, Andreas Fink, Evelyne Lutton, Penousal Machado,
Stefan Minner, Michael O’Neill, Juan Romero, Franz Rothlauf, Giovanni Squillero, Hideyuki Takagi, A. Şima Uyar, and
Shengxiang Yang, editors, Applications of Evolutionary Computing, Proceedings of EvoWorkshops 2007: EvoCoMnet,
EvoFIN, EvoIASP, EvoINTERACTION, EvoMUSART, EvoSTOC and EvoTransLog (EvoWorkshops’07), volume 4448/2007
of Lecture Notes in Computer Science (LNCS), pages 91–100, València, Spain, 2007. Berlin, Germany: Springer-Verlag
GmbH. doi: 10.1007/978-3-540-71805-5 10.

Distributed Computing Thomas Weise 28/28