WHITE PAPER
EDGE COMPUTING AND 5G
5G promises ultralow latency in transmitting large
volumes of data. But bottlenecks may occur if cloud
computing cannot respond fast enough. Processing
power at the edge of a network will be an essential
part in 5G delivering on its potential, and may provide
a new revenue stream for telecom providers.
By 2025, there will be more than 42 Figure 1: Distributing processing throughout the network
billion active IoT devices1 creating 90
zettabytes2 (1 zettabyte = 1 trillion
gigabytes) of data. An average person
will have more than 5,000 digital
interactions per day.2 Increasingly, Traditional cloud
APPLICATION HOSTING PLATFORM
these device and human interactions processing
will be in real time, through streaming Core/cloud data center
data, rather than intermittent OPERATING PROCESSOR STORAGE
SYSTEM
interactions, as is typically the
case today. Cloudlets/intermediate
data centers Distributed edge
In theory, this growth will all be computing at
enabled seamlessly by the new 5G Network components – different levels of
standard, which should reduce latency routers, switches, the network with
base stations processors,
by 50 times (from 50 milliseconds to 1
storage, application
millisecond) and increase bandwidth
platforms and
between four times, on average, to operating systems
Client devices
about 100 Mbps to 10,000 Mbps at
its peak.3
Source: Infosys
But as more rich media and time-
critical applications are deployed to prevalent, more passengers will Many standards and flavors of edge
benefit from this wireless network, the expect to use their travel time computing are evolving, based on
concern is that processing bottlenecks productively while on the move. This who is developing the technology and
will start emerging back at the data most likely will involve two-way rich what application it is being used for.
centers or in network nodes. media connectivity, either for video Some of those popular architectures
Consider facial recognition. For conferencing or online collaboration. being adopted today are multi-access
example, if the New York Police And as IoT devices proliferate and edge computing, fog computing
Department wants to use facial begin supporting “smart city” and cloudlets.
recognition for threat detection management, the reliance on large
in Times Square, it would need an
application that continuously captures
volumes of data to manage daily civil
tasks will become much more critical. Multi-access edge
images/videos of thousands of computing
moving people and objects, analyzes
the images and compares them with
Edge computing Multi-access edge computing,
or MEC, is an edge computing
existing data to identify suspicious The solution, at least in part, lies in
architecture developed by European
elements, and alerts the police. To be pushing more processing power away
Telecommunications Standards
effective, this complex process would from the core, and decentralizing
Institute, ETSI, which brings a cloud-
have to be done in a fraction of a throughout the network, closer to
enabled service environment closer
second. This might not be possible if the edge devices (Figure 1). This new
to the edge of the network. The idea
data has to travel through a choked evolution of “edge computing” will
is to bring computing close to the
physical network or compete for be crucial in helping 5G deliver its
user, but to not have it residing on the
resources at the data center. promise of ultralow latency.
user device.
The issue becomes more of a Historically, computing power has
This model creates three regions: the
challenge if, for example, during a swung back and forth between the
client or the user, the edge server
potential terrorist attack or another edge and the core, depending on
and the remote server. The MEC
emergency, the police want to quickly the technology architecture of the
“host” placed at the edge of the
shift their focus to a different part of time. Applications that benefit from a
network consists of an MEC platform
the city. This would suddenly move the lower-latency and higher-bandwidth
and resources to store and process
pressure to another segment of the 5G wireless network will start the
data.4 The MEC platform is a secure
network, which may not be ready for it. pendulum swinging toward a more
environment where services can be
powerful edge.
It’s not just emergencies either. As created and delivered. Low-latency
autonomous cars become more
External Document © 2019 Infosys Limited
Figure 2: The different layers within a fog computing architecture
Response time Storage time Area covered
Platform
Cloud Processor Minutes, Months,
Storage hours, days years Global
IoT app
Aggregation Processor
Seconds to Hours, days, Cities or
nodes minutes weeks wider area
Storage
IoT app Transient Local,
Milliseconds
Fog nodes Processor storage blocks
IoT devices IoT app
Sources: Infosys, Cisco
applications are offered or serviced stored there for a few hours and then Whenever a connected device has
by the edge server or MEC host while transmitted to the cloud for long- to transmit data, it offloads the data
complex requests are routed to the term storage.5 onto the nearest cloudlet (Figure 3).
remote server. It is interesting to The cloudlet stores and processes this
note that, in MEC, developers need
to clearly specify which application
Cloudlets data. It might decide to pass on the
data to the main data center if it sees
elements will run on the edge cloud or Cloudlet is a term coined by Mahadev doing so as being necessary. When the
MEC host and which ones should be Satyanarayanan and a team of connected device moves, the cloudlet
routed to the cloud. people working in Carnegie Mellon detects the next cloudlet closest to
University and Intel labs to describe the device and transfers all the data
Fog computing an intermediate component between
the edge and the cloud. Cloudlet, also
using defined protocols, essentially
orchestrating a cellular handshake.6
Fog computing is an edge computing known as a “data center in a box,” is a This ensures seamless low-latency
standard defined by Cisco. Fog mini cloud that has all the capabilities processing as the device travels.
computing too, like MEC, aims to of a cloud, but in smaller portions. It
All of these types of edge computing
bring the cloud close to devices, with has an operating system, storage area
will likely interact and work together.
a focus on IoT devices. Fog nodes are and a platform that hosts applications.
devices with storage, network and
computing capabilities, and can be
Figure 3: Cloudlets create mini clouds to manage mobile
deployed anywhere in the network as
shown in figure 2.5 Fog nodes nearest devices effectively
to the IoT device collect data from
the source. Then, based on the time
sensitivity and complexity of data, the CENTRAL CLOUD
fog IoT application directs the data to
the right processing station. The most
time-sensitive data is processed at the PLATFORM OPERATING APPLICATION STORAGE
SYSTEM
nearest free fog node. The decisions,
which can take a couple of seconds,
Handshake
are directed to the aggregation
Platform Application
nodes placed in the network, and the Cloudlet 1 Cloudlet 2 OS Storage
decisions that can wait are sent to data
centers for storage and processing.
The fog nodes also act as intermediate
storage locations. The data collected is
Source: Infosys
External Document © 2019 Infosys Limited
The industry is at a stage where Moving devices computing platforms — MEC hosts,
standards are still being set, and it fog nodes and cloudlets — will need
The end points and sources of data,
won’t be until 5G networks are much to communicate flawlessly with each
meaning the IoT devices, will not
more established that we see the types other and also with the core data
always be stationary. As these devices
of applications that are in demand. centers. The data has to be routed to
move, the system should automatically
This will determine the specific the right nodes in the shortest time, as
be able to connect them to the
architectures and approaches that low latency is the goal here.
nearest edge node based on the
are taken. location. If this is not done instantly, Also, as the user moves, the edge
As edge computing evolves, however, the data would travel to the previously node should do a handshake with the
it’s worth bearing in mind the many connected edge node, which might next-nearest node/cloudlet. All of this
now be far away. Then, the advantage requires a sophisticated controlling
challenges and complexities that need
of edge computing will be lost. This layer or control plane that can handle
to be overcome to scale this concept in
also has to be done perfectly to ensure this routing. Software-defined network
the 5G era.
a seamless user experience. architecture enables cloud providers
to centrally control the network
Device capabilities switching to find the most effective
Security risks
The source/end devices, substations route; this will have to be extended to
and routers all need to be equipped With increasing points of storage and accommodate edge computing. The
with the hardware and software processing, security vulnerabilities are control plane should create a virtual
capabilities required to support bound to increase. The network and system to classify data, balance the
edge computing. The hardware service providers will have to enhance load and distribute the traffic between
providers need to decide what kind security measures to address the the edge computing nodes. It should
of computing and storage capacities possibilities of malware attacks. As the also be able to identify the best
these devices will have. If the device network becomes more connected, possible node to which all the data can
has a lot of processing power, it would malware or security breaches can be transferred as the device moves.
be very fast. But it would be heavier, be contagious.
expensive and require more battery Costs and commercial model
power. Therefore, the hardware Intelligent edge Perhaps the biggest challenge is
providers would need to find a balance
between processing power and the Edge nodes need to be equipped overcoming the fact that to build
size of the device. with intelligent decision-making an edge network in 5G would incur
capabilities to ensure quick response significant costs. Certainly, the ever-
time. For example, in case of facial declining price of chips and storage
Switching nodes
recognition application, the edge will help. But the fact remains that, in
For a seamless user experience, the nodes should be able to analyze the the early days, the business case for
software needs to be designed to images to identify any suspicious upgrading to an edge network may
intelligently route the data to different activity. The edge node in an be difficult to prove, until it’s clear
processing nodes. For this, it will need autonomous car should be able to what applications can be monetized
to consider many parameters, such apply machine learning algorithms on the network. It does leave 5G edge
as the volume of data, current traffic to predict the behavior of other computing in somewhat of a “chicken
in the network, nearest edge node vehicles on the road dynamically and and egg” situation, whereby the value
that is free, the best path to route the move accordingly. can’t be proved without the network,
data, which part of the data is to be and the network can’t be built without
processed by an edge node and which proven value. That said, progress is
Distributed control
one is to be routed further, how critical occurring, and we expect that it will
the decision is, how time-sensitive There are two management challenges only take the emergence of one “killer
the decision is, and many more. The that this type of edge network will app” to quickly incentivize the industry
way software is architected and how create: managing the interconnections to invest.
it interacts with the operating system across the multiple infrastructure
of each device will be incredibly nodes, and enabling a virtualized Once we reach that stage, there
complex, and will require a virtualized layer for software developers to easily are many ways in which telecom
control layer to support developers create applications that can benefit companies can monetize edge
(see below). from a distrusted edge network. On computing. They can develop their
the infrastructure side, different edge own edge networks to facilitate
External Document © 2019 Infosys Limited
zero-latency use cases for their infrastructure as well as playing a partnerships with telecom companies.
subscribers — both consumer and part in the content and value that is The hyperscale cloud providers like
corporate. They can also develop a generated over it. However, with the AWS and Microsoft are already looking
virtual edge platform where software arrival of 5G edge computing, the to develop edge computing solutions
developers can build and sell new low- power balance could soon tilt away for 5G. They intend to partner with
latency, highly critical applications to from them. telecom companies for last mile
network subscribers. delivery. Microsoft recently announced
Today telecom carriers are the closest partnership with Telefonica to explore
Telecom companies can define a ones to the users and IoT devices, and intelligent applications and services in
management plane to handle the will also control the 5G infrastructure. the network.7
data flow and manage service level Edge computing and 5G provide
It’s clear that 5G and the potential
agreements. This could orchestrate a unique opportunity for telecom
creation of a powerful edge network
the entire flow of data between the providers to move up the value chain,
will start to reorder the industry.
devices, edge nodes and core, and after having lately been characterized
Many new opportunities will emerge
generate commercial value by being as “dumb pipe” service providers.
for all involved, from infrastructure
an intelligent layer between content This point has not been missed by
manufacturers to content providers
providers and consumers. the industry. Telecom companies
and software developers — and
have already started exploring
telecom providers need to embrace
opportunities edge computing on
Edge — a lifeline for 4G network. Telstra is looking at
this opportunity too. To ensure that
they do not miss the opportunity,
telcos? converting its base stations and towers
into computing nodes.
telecom carriers need to up their game
technically and commercially to truly
Cloud providers such as Amazon,
The cloud leaders see this threat position themselves as the smart
Google and Microsoft have been the
and have started investing in the “middleman” in the era of connected
power players in the cloud computing
ecosystem. They are looking for computing and 5G.
era, both in terms of owning the
References
1.
“IoT Signals report,” Microsoft, July 2019, [Link]
locked-by-addressing-skills-shortage-complexity-and-security/
2.
“IDC: Expect 175 zetabytes of data worldwide by 2025,” Network World, December 2018, [Link]
cle/3325397/[Link]
3.
“5G technology needs edge computing architecture”; Cisco; [Link]
[Link]
4.
“MEC software development”; ETSI; February 2019; [Link]
[Link]
5.
“Fog Computing and the Internet of Things”; Cisco; 2015; [Link]
[Link]
6.
“Cloudlets: at the Leading Edge of Mobile-Cloud Convergence”; CMU; 2014; [Link]
[Link]
7.
“Telefónica and Microsoft establish strategic partnership”; Microsoft; February 2019; [Link]
ca-and-microsoft-establish-strategic-partnership-to-design-the-telco-of-the-future/
External Document © 2019 Infosys Limited
Authors
Gnanapriya C Samad Masood Rachana Hasyagar
Associate Vice President Infosys Knowledge Institute Infosys Knowledge Institute
gnanapriyac@[Link] [Link]@[Link] [Link]@[Link]
About Infosys Knowledge Institute
The Infosys Knowledge Institute helps industry leaders develop a deeper understanding of business and technology trends
through compelling thought leadership. Our researchers and subject matter experts provide a fact base that aids decision
making on critical business and technology issues.
To view our research, visit Infosys Knowledge Institute at [Link]/IKI
For more information, contact askus@[Link]
© 2019 Infosys Limited, Bengaluru, India. All Rights Reserved. Infosys believes the information in this document is accurate as of its publication date; such information is subject to change without notice. Infosys
acknowledges the proprietary rights of other companies to the trademarks, product names and such other intellectual property rights mentioned in this document. Except as expressly permitted, neither this
documentation nor any part of it may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, printing, photocopying, recording or otherwise, without the
prior permission of Infosys Limited and/ or any named intellectual property rights holders under this document.
[Link] | NYSE : INFY Stay Connected
