1 Computing Paradigms

1.1 INTRODUCTION TO COMPUTING

One of the astounding scientific developments of humanity is the Computer. The
word comes from the Latin word “Computare” that consists of two parts – “Com”
meaning “together” and “putare” meaning “To settle (an account)”. Computers
are the devices used for computing. “Computing” is drawn from the Latin word
“Computatio” that means reckoning or calculation. The process of computing
uses computer technologies to complete a task. The process of Computing in­
cludes activities performed using computational devices to manage, process, and
communicate information.
We have moved from the era where computers were considered as a luxury
to an era where it is considered as necessity. We use some form of computing
in our daily life without much ado. Using a cell phone, washing machine,
communicating using email, financial transactions all include some form of
computing. Task automation triggered using computers increases accuracy while
reducing manual work, power, cost, and time. As quoted by the Association
of Computing Machinery (ACM) [1], “Computing and computer technology
are part of just about everything that touches our lives from the cars we drive,
to the movies we watch, to the ways businesses and governments deal with
us”. Computing drives the innovations behind various fields like sciences,
engineering, business, entertainment, and education.
The urge of humanity to make a device capable of solving complex problems
between 1600 AD and 1800 AD leads to the invention of mechanical calculators.
Blaise Pascal invented the first mechanical calculator in 1942. In 1671, Gottfried
Wilhelm von Leibniz, the German mathematician designed the mechanical calcu­
lating machine called the Step Reckoner based on Pascal’s invention. World War II
saw the rise of electro-mechanical hybrid calculators. The first calculating computer
Electronic Numerical Integrator and Computer (ENIAC) was completed after the
war in 1946. It was basically designed during the war to handle complex calcula­
tions connected with the movement and strategies for the military during the war.
Many scientists, mathematicians, and computer technicians worked hard to make
the modern computer a reality. The computational capacity represented by the speed
of the computers doubled every 1.5 years between 1975 and 2009. Beyond this
sequential computing gave way to parallelism. Details are given in Section 3.
The growth of Internet, digital devices, and computer technology is the backbone
behind a number of disruptive revolutionary business innovations like video
streaming, mobile applications, retail medical clinics, Uber, Swiggy, digital cam­
eras, and online education. In this digital age, computing has infiltrated into a
number of domains as given in the following section.

DOI: 10.1201/9781003230946-1 1
2 Edge Computing

1.2 THE MAJOR IMPACTS OF COMPUTING

Technologies that trigger innovations ensure the liveliness of a society. Bright and
ethical technologists are the driving force behind the usage of computing for the
benefit of the society. Today computing is an essential tool for every business,
banking, government, entertainment, industry, education, and administration. It can
be said of all large organizations, whether the department government or private,
use a computer for a variety of their daily business and it is the fastest-growing
industry in the world today [2].

• Business: Businesses use computers for a variety of purposes. Computing can

be used to analyze current business trends, predict sales of a product, identify
inventory, predict target customers, calculate salary, issue/receive orders,
communicate business letters, create invoices, and more. They are also an
essential tool for office automation that includes create/edit reports/docu­
ments, visualize sales, create presentations, manage business data, and more.
Internet has fostered the growth of E-commerce industry.
• Finance and banks: Computing has fostered the growth of cashless so­
ciety. The use of computers provides effective, fair, and personalized
banking solutions. It can be used for maintaining accounts, managing in­
vestments, analyze profitability, and hence aid in reducing the operational
budget of banks.
• Industries: Computers are used extensively to control and monitor the pro­
duction process in various industries. This includes management of employee/
customer information, maintaining and analyzing production and sales
information, scheduling production, and so on.
• Healthcare: In the field of medical science, computers are used extensively
to manage patient details, schedule appointments, administer purchase and
inventory of medicines/equipment, aid in disease diagnosis, and for medical
research. Advancements in the computing field have enabled the blind to see,
the deaf to hear, the dumb to communicate, improvise learning, administer
medicines among a few.
• Home: Smart home systems using microcomputers enhance home security
and enable users to control various devices like air-conditioners, lights, tel­
evisions remotely. Online shopping and bill payment systems simplify var­
ious tasks. Spare time can be used for other purposes. Computers also help in
budget planning and maintaining inventory. Children can use online learning
resources to their benefit. Thus computers have turned into machines to
maintain, manage, and communicate information in this modern world.
• Education: Computer technology can change the way we learn. Blended
learning that combines classroom learning with online resources can enable
students to understand the concept better. The use of videos, streaming, and
augmented reality demonstrate a concept with animations and hence improves
the learning of the student. Virtual classrooms can improve interactions.
Teachers and management can use computers to store, maintain, and analyze
student information. State of art distributed ledgers like blockchains help to
Computing Paradigms 3

maintain student information in a secure manner. World Wide Web (WWW)

is a huge repository of information. Computers facilitate students to search
and access information from this huge repository. Computers aid in student
registration, class scheduling, processing of examination results.
• Law: Computers are used extensively to gain access to information and for
documentation. Intelligent systems can be also used for aiding lawyers for
diagnostic and decision-making purposes. It includes analyzing and proces­
sing information to generate weekly and monthly reports, recording and
maintaining payments, maintaining a legal information diary on court pro­
cedures with search facilities, maintaining a record of users.
• Government sector: uses the field of computing for effective administration
and record keeping. Computing platforms can aid in effective storage/re­
trieval of data of citizens, services, projects, properties, organizations, and so
on. It can be used for effective planning, communication, and decision-
making. It can be used for traffic control, monitor enemy movement in bor­
ders, forecast weather, and predict disasters among a few.
• Entertainment: Computers are used extensively in various sports to predict and
improve the performance of players, games, play music, drawing, and so on
• Agriculture: Computers are extensively used by farmers in the agricultural
fields to record various soil and weather parameters and to automate various
agricultural tasks. Various portals provide guides to farmers based on the
daily conditions. Yield prediction, weather forecasts extensively rely on
computing.

Rapid development of science and technology has changed the business systems in
various domains. Manual tasks have been automated using computers. This reduces
risks and improves effectiveness of the business systems in various domains in­
cluding financial, industry, education, and administration. To conclude computers
have become the brain behind the society.

1.3 PARALLEL COMPUTING

Since 1960, the fundamental components of a computer as described by the tra­
ditional Von Neuman Architecture consist of the memory, central processing unit,
and the arithmetic unit and logical unit. Moore’s law states that the transistor
density doubles every 1.5 years. This in turn doubles the processing power. But this
also reduces the space between the components and increases the power dissipation.
This is because power dissipation increases by square of increase in clock speed. On
an average, processor performance is increased by 52% every year between 1986
and 2002. After 2002, the performance stabilized. Clock speed could not be in­
creased further due to power dissipation. Moore’s law on increasing transistor
density continues but clock speed has to be stabilized. This resulted in the emer­
gence of a parallel computing paradigm using multiple cores. Here instead of in­
creasing transistor density in a chip to improve performance of one processor, those
transistors can be used to pack multiple processors into the same chip. Thus
sequential computing gave way to parallelism.
22 Edge Computing

• Here the entire set of resources in a VO provides a single system image to

the user. Many resources are virtualized as one. A job is split among many
under-utilized resources. This might not be beneficial from the business point
of view.

1.8 CLOUD COMPUTING

Cloud computing is a form of utility computing just like grid, where users pay for
the computing resources used by them. Cloud [9] is defined as a parallel and dis­
tributed system consisting of a shared pool of abstracted, virtualized, dynamically
managed computing resources (like storage, computing, platforms, network, ap­
plications) that are maintained in data centers. The resources are dynamically
provisioned and reconfigured based on demand from the user. Hence resource
management in cloud is elastic in nature. This is done automatically with less
manual intervention. Hence services are abstracted from the user. SLA acts as a
contract between the service provider and consumer. Users are then charged based
on their usage of the computing resources. This forms a business model. From
business point of view, this model reduces the capital and operational expenses of
an organization.

1.8.1 CHARACTERISTICS OF CLOUD ENVIRONMENTS

Cloud computing models enable ubiquitous, convenient, on-demand network access
to a shared pool of configurable computing resources. The following are the key
characteristics of a cloud computing environment:

• Service-oriented architecture: All functionalities are offered to users as ser­

vices through interfaces. These services can be easily extended, composed,
orchestrated to offer the desired functionality to the user. SOA facilitates
loose coupling and abstraction. Services can be customized to meet the user
needs.
• On-demand self-service: Services can be provided dynamically, without
human intervention as and when required by the user.
• Broad network access: The cloud provides services to heterogeneous client
devices such as mobiles, laptops, and workstations through a standard net­
work like the Internet.
• Resource pooling: The provider’s resources are pooled and allocated/de-
allocated to users dynamically, on-demand.
• Rapid elasticity: The resources in the cloud are dynamically provisioned and
released on-demand as and when required by the application.
• Measured service: The services provided by the cloud can be metered and
monitored.
• Location independence: Users access the cloud services anywhere and any­
time through the Internet.
• Business Model: Cloud computing provides a business model for the IT
Computing Paradigms 23

industry where the computing resource usage can be monitored. Payment can
then be based on the usage of resources.
• Virtualization: This technology abstracts the hardware/software resources
from the user. One physical resource is compartmentalized into many virtual
resources that can be offered to various users. This has resulted in multi­
tenancy where a single resource is utilized by many users. This in turn in­
creases resource utilization and reduces the cost.

1.8.2 CLOUD MODELS

Cloud computing models are categorized based on their delivery of services or
based on their deployment model. These two categories are discussed in the fol­
lowing section.

1.8.2.1 Cloud Services Models

The cloud services model defines the core functionality of the services implemented
at different levels and delivered to the clients. Such delivery models (Fig. 1.17) can
be categorized as follows:

• Infrastructure as a Service (IaaS) allows providers to set up and deliver

virtualized hardware resources such as storage, compute, and networking to
the users. Communication as a Service and Data storage as a Service are also
some examples of IaaS. IaaS falls under the category of utility computing.
• Platform as a Service (PaaS) offers developers with the tools, platforms, and
environment to build and deploy their applications on the cloud. It includes
platforms for the development and deployment of services, environments for
runtime execution, databases, APIs, etc. This is also termed as framework
computing.

FIGURE 1.17 Cloud service models.

24 Edge Computing

• Software as a Service (SaaS) offers applications operating in the cloud that

can be accessed remotely on-demand as a service by the end-users. Users
are thus freed from the burden of installing and using the required software
locally. It is termed as business computing.

Following this, other services, such as database as a service and analytics as a

service, can also be set up. Big Data as a Service (BDaaS) maintains and manages
voluminous data on the cloud storage and offers continuous access of the same to its
owner.

1.8.2.2 Cloud Deployment Models

Based on the functionality and accessibility of cloud services [9], the cloud offers
four deployment models (Fig. 1.18):

• Public cloud: Users access the resources including infrastructure, frame­

works, applications, and services provided by the service providers over the
Internet. Users subscribe to the services they access and have to pay based on
their usage.
• Private cloud: This type of cloud is in-premise and services are offered over
the intranet. It is more secure when compared to a public cloud.
• Hybrid cloud: It is composed of multiple public/private clouds. This model
offers the advantages of public and private deployment models.
• Community cloud: This can be a public/private cloud shared by a set of or­
ganizations under the same administrative domain.

NIST architecture of cloud is layered and is based on the essential characteristics of

cloud at the top, followed by the delivery models and deployment models. The
following are the advantages of the cloud model of distributed systems:

FIGURE 1.18 Cloud deployment models.

Computing Paradigms 25

• Cost reduction: It reduces the Capital expenditure of purchase of equipment

and operational expenditure of maintaining the cloud environment in an or­
ganization. Further, the operational expenditure including power, wages, and
upgradation is shared by all the users.
• Better flexibility and scalability as new infrastructures can be added on the fly
without affecting the services being offered.
• Better resource utilization as the same resource is shared virtually among
many users.
• Data backup and disaster recovery are handled automatically by the service
providers.
• Cloud resources can be used anywhere anytime using any access technology
by the users.
• Cloud resources are highly available as another instance can be started on
failure of the existing instance.

Some of the challenges faced in cloud computing include:

• Reliable Internet Connection should be available to access the resources.

• Security of customer’s data is managed by the service providers.
• Dependency on cloud service providers/vendors.

1.9 OTHER COMPUTING PARADIGMS

1.9.1 UBIQUITOUS COMPUTING
This section compares ubiquitous, pervasive, and mobile computing. Desktop
computing enables a user to carry out computing functionalities using a single
desktop. Here a single desktop is used by a single user. Ubiquitous computing
moves to the era when invisible resources are used by many users. It is a many-
many relationship. Ubiquitous computing, which means "omnipresent". Cloud is a
ubiquitous environment as users can access the remote resources hosted in the cloud
using Internet and various access technologies. Even though resources do not move,
these resources can be accessed from any location using Internet. Hence mobility is
high. A pervasive computing environment embeds computing in various devices
like IoT. The resources are much smaller when compared to that of mobile systems.
Mobility of such systems is low. They are used in applications like smart homes.
Mobile computing embeds computing in devices carried by the user. Mobile
computing involves mobile communication, mobile hardware, and mobile software.
The mobility of such systems is high. Computing using cell phones comes under
this category. The main characteristics of ubiquitous computing are:

• seamless access to remote information resources and fault-tolerant communication

• when compared to pervasive and mobile computing, the user does not possess
the resources and hence does not have control over the remote resources.
• The user can access the resource anytime, anywhere using suitable access and
communication technologies.
 2 Edge Computing and
Its Essentials
2.1 INTRODUCTION
Consider a smart building with video cameras to detect the movement of people.
Video signal is continuously streamed to the cloud server. The motion detection
application in the cloud server can detect features and store to the database. As large
volume of video data is streamed continuously to the cloud, significant network
bandwidth is consumed. The cloud server should also analyze the video footage
from all the cameras simultaneously. These issues can be overcome if the motion
detection application could be executed close to the cameras. Only the required
clips can then be transferred to the cloud for further analysis and storage.
Traditionally, real-time data collected from various sensors is propagated uplink
to cloud servers for further processing and storage. The cloud infrastructure is
highly suitable to process such large amount of continuously changing hetero­
geneous data also termed as big data. But, big data processing in distant cloud
infrastructure fails when data sources are distributed across multiple locations and
processing is to be done near real-time. This motivates the need for an alternative
paradigm that is capable of performing computations closer to the sensors or data
sources. Such an infrastructure located at the edge of the network is capable of
performing low latency computations, aggregating the data, and then transferring it
to the cloud for further analysis. Also, downstream responses from the cloud are
returned to these nodes. The edge of the network can be a room, a building, or a
campus based on the network coverage.
It becomes essential to process voluminous data generated by IoT devices in
real-time in proximity to data sources using edge computing infrastructure. This
not only reduces latency in computation and usage of networking resources but
also enforces data security. This section discusses the evolution of edge com­
puting paradigm through the technology preparation phase, rapid growth phase,
and stabilization phase [1].
Technology preparation phase started off with content development networks
(CDN). CDNs with geographically distributed caching servers enabled users to
access the nearest server (1998). Network traffic and latency were decreased. While
CDN emphasizes on data caching, edge computing focuses on function caching. In
2005, function caching was applied for personalized mailbox management to reduce
latency. Cloudlet (2009) was used to downstream cloud information to network
edge. Edge data processing initially started with mobile devices (2010). Mobile
edge computing (MEC) emphasizes the establishment of edge servers between the
cloud server and edge devices for offloading computations. The concept of fog
computing was fostered by CISCO (2012). Fog computing focuses on a virtualized

DOI: 10.1201/9781003230946-2 39
40 Edge Computing

computing platform to migrate cloud tasks to the network edge. While fog com­
puting aims at optimizing communication at the infrastructure level, edge com­
puting (2013) focuses on computing and network resources at the device, edge
server, and cloud server layers. Cloud-sea computing emphasizes on the cloud end
and human/physical world end of the system. But, edge computing concentrates on
the communication path (upstream and downstream) between the cloud and the IoT
devices.
Since 2015 edge computing has attracted more attention from industry and
academia. In 2015 edge computing gained its formal definition from “Edge com­
puting: Vision and challenges” white paper. The paper defined edge computing as
an enabling technology allowing computation to be performed at the edge of the
network, on downstream data on behalf of cloud services, and on upstream data on
behalf of IoT services. In 2016 NSF in collaboration with Intel established in­
formation center networks in wireless edge networks (ICN-WEN). NSF Workshop
on Grand Challenges (2016) and edge computing symposium in China (2017)
exposed the technology to the scientific community. LinuxEdgeX – the common
open framework for IoT edge computing was built in 2017. The year 2018 saw
active industrial participation in development of edge computing platforms.
Automotive Edge Computing Consortium was formed in 2018. The first book on
edge computing was also published in this year. Kubernetes used in both cloud and
device environments was developed collaboratively by Cloud Native Computing
Foundation (CNCF) and Eclipse Foundation in 2019. The usage of edge computing
in healthcare domain was focused in the Bio-World Conference and Expos (2019).
This rapid growth trend is expected to stabilize in 2021.
Edge computing analyses data by harnessing the compute power of the resources
outside of traditional data centers. These computational resources are located in
proximity to the data sources. Data caching/storage, data processing, service de­
livery, IoT management, and privacy protection are some of the services offered by
this layer. Edge computing environment up-streams data from devices to cloud
servers and down-streams data from cloud servers to the devices.
When cloud computing was introduced, all the computations were shifted to the
cloud. Due to network latency and the transmission cost, it is more logical to
perform related tasks at the network edge. With the improvement of the processing
power and capability, the amount of tasks performed on the edge will continue to
grow [2]. The major advantages of edge computing include:

• Reduction in computation latency to meet the customer demands.

• Reduction in network bandwidth.
• Preservation of data security and privacy.
• Good reliability even on network failures.
• Reduction in operational costs due to communication, storage, and
processing.
• Better data governance due to improvement in quality and usability of data.
• Improved scalability of cloud by computation offloading to the network edge.

The major components of an edge computing system [3] (Fig. 2.1) include:
Edge Computing and Its Essentials 41

Cloud Layer

Internet

Edge Layer

Network
Layer

Perception
Layer

FIGURE 2.1 High-level architecture of an edge computing system.

Perception layer: It consists of the edge devices. An edge device is a special-

purpose cost-effective hardware designed to perform a specific task effectively. The
edge device has limited compute/storage resources. Some examples of such devices
include transducers, sensors, actuators, logs, and cameras that perform functionality
of gathering and/or transmitting data. Some edge devices have processing power to
do additional activities. Analytics applications on image, video, text data gathered by
these devices need to be deployed and managed based on the resource availability [4].
Networking layer: It is responsible for connecting devices, edge systems, and
cloud systems. It comprises various communication and data transfer protocols.
Edge computing layer: The major components in this layer include the edge
server and gateways. An edge server has higher computational/storage capacity
when compared to an IoT device. Workload that cannot be carried out in a resource-
constrained edge device can be done using these general-purpose compute nodes or
a cluster of nodes. Edge gateways in this layer perform specialized network func­
tions including protocol mapping, network termination, and tunneling and firewall
protection.
Application or processing layer: This layer is responsible to carry out complex
data-intensive tasks and to store voluminous data. Private or public cloud servers in
this layer act as repositories for sharing data among the nodes. Cloud servers
perform complex data-intensive/compute-intensive applications that cannot be
hosted in edge servers. Cloud servers can also be used for resource allocation and
job management at the edge nodes.
Apart from the above-mentioned components, a suitable interface is to be pro­
vided for the users to monitor, manage, and interact with the system. The next
section will provide a detailed architecture of the edge computing system.
42 Edge Computing

2.2 EDGE COMPUTING ARCHITECTURE

By considering edge and cloud computing paradigms as complementary, a synergy
can be reached between them. Edge servers can handle latency-aware computing in
close proximity to data sources. Shared global computation/storage, management,
and data privacy can be handled effectively in the cloud server. Fig. 2.1 discussed
the various layers of a generic edge computing model. Fig. 2.2 provides a more
detailed view [5] considering a network of edge servers.

2.2.1 EDGE DEVICES

Devices that collect data in the site including sensors, cameras, microphones, re­
corders, web monitors, data loggers can function as edge devices. Edge devices can
also be actuators and visualization tools. Objective of the edge device is to collect
information and send upstream or transfer information downstream for actuation.
Edge devices have usually limited compute and storage capacity. Sometimes they
can have higher capacity to do additional functionalities. Edge devices contain
transducers that convert physical signals to electrical and vice-versa.

2.2.2 EDGE SERVER CLUSTER

When a distributed peer-to-peer connection of edge servers is considered, some
edge servers may be close to a particular edge device while others may be far
away. Computation and storage are replicated in multiple edge server nodes. The
edge servers can be present at the device edge or at the network edge. Some
workloads can be processed at the device edge maintained close to the sensor/
actuator. Other workloads have to be migrated across to other edge servers called
as local edge servers, where they can be processed. These servers may be farther
away from the device but within the network edge. Workloads include both ap­
plication workload (that includes compute and storage) and network tasks. As the
local edge contains more resources they can perform complex computing and
networking functionalities.

FIGURE 2.2 Edge computing – a detailed architecture.

Edge Computing and Its Essentials 43

Device edge nodes only perform simple functionalities. The compute nodes
perform less compute-intensive operations. The gateway nodes additionally per­
form simple networking functions from the device to the local edge. Network la­
tency involved may be in microseconds. Communication protocols between the
devices and between the device and the local edge work over shorter distances
and consume low power. Edge devices with additional functionalities act as device
edge nodes.
Local edge nodes have additional compute, storage, and network capability.
Such edge nodes can be used for computation. A small-scale data center (cluster of
nodes) at the network edge called the cloudlet can be used. Such a cloudlet can store
data at the network edge, host intelligent applications (like video/text analytics) to
process data, and filter data to be transferred upstream to the cloud. Network latency
may be in tens to hundreds of milliseconds.
Network function virtualization handles networking functions (like load balan­
cing, security) with software-based technologies (called virtual network functions
(VNFs)) rather than hardware. VNFs decouple software from the underlying
hardware. This offers flexibility to vendors to create a more scalable and agile
network using software functions. Some of the networking functionalities provided
by VNF at the local edge include routing, firewall, Radio Access Network (RAN),
Evolved Packet Core (EPC) framework for a 4G Long-Term Evolution (LTE)
network. Protocols used between the edge server and the cloud have large
throughput and high speed. Data caching and filtering, visualization, real-time
computation, and control are common tasks done at the edge level. Resource
allocation and management is a major issue. Usually, cloud servers handle it.

2.2.3 CLOUD SERVER

The server is used to perform complex computation, resource allocation and
management, data summarization, and secure data storage. Resource management
defined by software framework in the cloud offers middleware services to manage
the resources of the entire system and maintains quality of service for the appli­
cations. These services aim to balance application latency and cost by choosing a
suitable resource for a task either in the device, edge server, or cloud server layer.
Some of the services offered in this layer include:

• Data management service is used to access information stored in various SQL

and NoSQL data stores.
• Monitoring service keeps track of the current status of applications and re­
sources in the entire system (edge and cloud).
• Knowledge base maintains past information on resource/application demands.
• Profiler creates resource/application profiles based on historical information
in the knowledge base and as well as current information obtained from the
monitoring services.
• Performance prediction service forecasts the resource utilization and perfor­
mance based on the historical information on resource demands. This in turn
affects the application performance.
44 Edge Computing

• Resource provisioning service allocates both cloud and edge resources

dynamically for various tasks based on the information provided by the
monitoring, profiling, prediction services, and user requirements.
• Task placement service aims at identifying a suitable resource at the device,
edge, or cloud layer to execute the task effectively. It uses the information
from the resource provisioning service.
• Security service provides the required authentication, authorization, and
confidentiality for various applications.

Consider a use case of a fire detection system using drones. Activities done in
different layers of the system include:

• Video analytics software to detect fire and network functionality for the
application is deployed in the device edge and local edge.
• Drones capture videos. Initial video processing (like color extraction) may be
done in the drones and device edge.
• Captured frames are sent to local edge for object detection (a more complex
functionality).
• Frames after object detection are transferred to cloud for further management.
In cloud, data from multiple drones may be analyzed to detect an activity and
report.
• If retraining is required in device or local edge, a new model may be trained
in the cloud and uploaded in the network edge.

The following sections discuss devices and networking protocols commonly used
for edge computing.

2.3 BACKGROUND ESSENTIALS: IOT DEVICES

Smart objects in Internet of Things system need to be context-aware and interact with
the environment. So, environmental attributes are obtained by these smart objects
through sensors. They also cause actions to the surrounding environment using ac­
tuators. Transducers convert signals from one form to another. Sensors and actuators
are transducers. IoT sensors are mostly small in size, cost-effective, and consume less
power. They are energy-constrained. Some of the common sensors are listed:

2.3.1 MOBILE PHONE-BASED SENSORS

Smartphones have inbuilt sensors and data processing capabilities. Smartphones can
be used to sense various physical elements.

• Movement patterns like walking and running can be sensed using accel­
erometers. Mechanical, capacitive, and piezoelectric accelerometers are
common.
• Orientation can be sensed using gyroscope.
• Visual and audio information can be collected using camera and microphone.
Edge Computing and Its Essentials 45

• Magnetic fields detected using magnetometer can be used to find direction

and presence of metals.
• Location of the phone can be detected using global positioning system (GPS)
• Intensity of the ambient light can be sensed using light sensor.
• Distance/proximity can be measured using the proximity sensors that use an
infrared (IR) LED, which emits IR rays.

Sensors to measure temperature, atmospheric pressure, and humidity like ther­

mometers, barometers, and humidity sensors may also be available in smartphones.
Some of the applications that use sensor data and computing capabilities of
smartphones include:

• Activity detection using camera, microphone, accelerometer sensors from

smartphones.
• Fitness applications that use sensors in smartphones to track physical activ­
ities, diet, exercises, and lifestyle.
• Analyzing smoking pattern and mental health can also be done using sensors
and computing capabilities of smartphones.

2.3.2 MEDICAL SENSORS

Wearable devices can remotely monitor the health of a patient using medical
sensors to measure the heart rate, pulse, blood pressure, body temperature, re­
spiration rate, and blood glucose level. Wearable devices in the form of smart­
watches, wristbands, monitoring patches, and smart textiles can be used to monitor
the vital health parameters of the patient. Data measured from these sensors should
be related to the contextual information to draw correct inference.

2.3.3 NEURAL SENSORS

These sensors use electroencephalography (EEG) to read various brainwaves in­
cluding alpha, beta, gamma, theta, and delta waves. Based on this the brain activity
can be deduced to manage stress, for improved learning and mental health.

2.3.4 ENVIRONMENTAL AND CHEMICAL SENSORS

These sensors are used to obtain environmental parameters like temperature, hu­
midity, pressure, water pollution, and air pollution. The electronic nose (e-nose)
and electronic tongue (e-tongue) sense chemicals based on the odor and taste.
These sensors can be used to monitor the pollution level and quality of food and
agricultural produce.

2.3.5 RADIO FREQUENCY IDENTIFICATION

Radio frequency identification (RFID) is an identification technology. It has a small
chip called the RFID tag along with an antenna. Ambient changes in magnetic field
46 Edge Computing

are encoded as signals which are then transmitted. These transmitted signals can be
read by an RFID reader. When the RFID tag is attached to the object to be tracked,
the reader detects and records its presence when the object passes by it.

2.3.6 ACTUATORS
These devices cause changes in the environment by converting the electrical energy
into other forms. They include heating or cooling elements, speakers, lights, dis­
plays, and motors. In a smart home system, actuators can be used to lock/unlock the
doors, switch on/off the lights, control temperature, and notify users.
Consider an operating environment of a mining industry. Such an environment is
hazardous. It is essential to monitor the employees closely during their working
hours. Sensors can be used to monitor the health of the employees as well as the
status of the work. Environmental sensors to detect poisonous gas, ambient tem­
perature, and lighting can be used to warn the employees. Vital signs of the em­
ployees can be collected and monitored to find if they are in danger. These sensor
details can be analyzed in the local edge at the operation center to ensure the safety
of the employees. Edge computing is thus highly beneficial to monitor the safety of
the miners.
This section discussed the perceptual component of the edge computing en­
vironment. To communicate between various components of the edge computing
system an effective communication environment is essential. Section 4 discusses
the common network and data protocols that can be used in an edge-computing
system. Section 5 discusses the network management using Software-defined
networks.

2.4 NETWORKING ARCHITECTURE

IoT devices capture a lot of real-time data in heterogeneous formats. These have to
be communicated using suitable communication protocols to the nearby gateways
for further processing and transfer of data. The communication protocols can be
categorized into two types. The first set consists of network and data link layer
protocols [6]. The second category is the data transfer protocols that can be used in
the application and session layers (Fig. 2.3). Some of the commonly used network
layer protocols [7] include:

• Bluetooth: is a secure, short-range, low-power, low-cost, wireless transmis­

sion protocol between electronic devices. Data exchange between electronic
devices over short distances is facilitated using 2.4 GHz wireless link.
Bluetooth star, mesh, and point-to-point topologies are highly suitable for
smart home applications.
• Wi-Fi: is a family of wireless network protocols, based on the IEEE 802.11
family of standards, which are commonly used for local area networking of
devices and Internet access. Any electronic device within range of a wireless
modem can attempt to access the network. Sensors in healthcare and smart
home applications commonly communicate with the gateway using Wi-Fi.
 6 Edge Computing Use
Cases and Case Studies
6.1 USE CASES
Every organization must have a clear goal of data storage and its architecture
to upkeep the planned use cases according to the priority/preference level.
Organizations are required to analyze gamut of highly prospective use cases and
its tendency to entice other stakeholders. Few sectors are described in this chapter
that best utilize edge computing systems by highlighting the opportunities and
point to ponder. Manifold ways exist to characterize the use cases and this chapter
is limited with most commonly applied domains/sectors/applications. Healthcare
sector case studies are presented elaborately in this chapter with a detailed
illustration and working module.

6.2 EDGE COMPUTING HIGH-POTENTIAL USE CASES

High-potential use cases that drive the real world are elaborated in this section [1,2].
Few of the highly prospective use cases are discussed as follows:

• Autonomous vehicles
• Smart cities
• Industrial automation
• Network functions
• Gaming
• Content delivery
• Financial sector
• Augmented reality/virtual reality
• Healthcare sector

6.2.1 AUTONOMOUS VEHICLES

Many automotive industries started to invest dollars of money in a smart way in
driverless cars using contemporary technologies. Though complete dependency on
machines is not possible, operating safely is a better solution. To function effi­
ciently, these vehicles are required to collect and analyze data from different or­
ientations, surroundings, and climatic conditions. It should have the potential to
decide on its own based on the previous experience/learning. In addition, present
feedback/decisions taken by the vehicles need to send back to the nearest edge
servers placed by the providers/manufacturers in order to operate effectively and to
communicate with the local government bodies. The efficacy is based on the data

DOI: 10.1201/9781003230946-6 139

140 Edge Computing

Cloud Server Ethernet/Fiber Optic

Ethernet/Fiber Optic

4G/5G 4G/5G

5G

FIGURE 6.1 Autonomous vehicle communication.

collected and analyzed for providing alerts and communicating the information with
nearby vehicles in the network with no inconsistency. The influx of data must be
transmitted to the vehicles in the network with no lack in bandwidth and latency to
offer high reliability in the edge-enabled environment. Edge-enabled autonomous
vehicles must have the capability to learn things from past decisions and to make
spontaneous onboard decisions using machine learning techniques like reinforce­
ment learning. Autonomous vehicle functioning is shown in Fig. 6.1 for better
understanding.

6.2.2 SMART CITIES

A massive amount of data is influx in Metropolitan cities from various sources such
as individuals, buildings, vehicles, and devices with the help of sensors. The data
are gathered, processed, analyzed, and managed for gaining insights and using the
resources efficiently, thereby improving the functioning of the city in traffic, ne­
cessary home utilities, trash, anomaly detection, and other societal services.
Existing schemes are deficient in providing an instantaneous response for taking
immediate actions. As a repercussion, edge emerges to offer solution to things
without any delay in network outskirts. Edge computing is a real-time permanent
solution for this changing environment receiving requests from multitude of things.
It provides a spotlight of people living condition and service utilization in me­
tropolitan places. In the future, the connection among the devices/people/buildings
will be proactive and reactive in nature. However, the level of collaboration among
the people/devices may vary according to the situation. Multitude of things and
intelligence merge together to make the place highly digitalized and interactive to
live life in a better manner. Manifold initiatives are launched by companies like
Edge Computing Use Cases and Case Studies 141

Google, IBM, and Cisco to support the government in this project. For instance,
IBM launched its smart city challenge to develop the urban ecosystem. This
strategy is well adopted by many countries across the world to improve the lifestyle
by focusing on e-services, sustaining novel ecosystems with high broadband net­
works. Universities like Tel Aviv join hands with this new venture with their focus
on smart transportation by incorporating artificial intelligence with the ecosystem.

6.2.3 INDUSTRIAL AUTOMATION

Manufacturing industries are the prime benefiters of the fourth industrial re­
volution that is popularly known as Industry 4.0. With the tremendous growth of
IoT devices and their management, the current generation is on the cusp of the
fifth industrial revolution. Integration of data storage facilities and computing in
industrial sector realizes the benefits such as minimum energy consumption and
cost. The flattened edge architecture realizes energy efficiency and predictive
maintenance in industrial sectors that in turn results in high production. Acquiring
data from various sources and effective analysis aids the industries to become
smart by increasing the production with customization as per customer require­
ments. In case of companies operating with low/poor bandwidth, edge computing
is considered as an irreplaceable solution. Edge computing is a more suitable
framework for industrial applications such as aircraft/missile launch, offshore oil
rig, energy manufacturing, fraud detection, monitoring of logistics equipment, and
so on. For instance, during missile launch edge framework must collect, analyze,
and monitor data on various weather factors without depending upon remote data
centers. The intelligent machine learning algorithms learn from past experience
and process the data in judicious manner that ultimately results in improving the
operational efficiency.

6.2.4 NETWORK FUNCTIONS

Due to the immense explosion of data and devices, every device is associated with
an IP address for identity, communication, and traceability. Current network
function is not suffice to upkeep the process efficiently as the interconnected
devices consume large bandwidth. Edge gateway is considered as the best solution
to reduce bandwidth consumption by placing it very close to the devices.
Rationalizing the data at edge gateway enhances the effectiveness of overall
function, especially bandwidth consumption. In contrast, improvisation processes/
procedures must be applied on routers, firewalls, and switches to uphold the
function either physically or virtually as it runs in cloud systems. Adaptation of
these devices to the local actuation point, edge remains as a highest challenge as
the rudimentary functions like security operations/rules and packet forwarding
must function on onboard without outsourcing to cloud to improve the overall
efficacy and performance. Many industries and researchers initiated work in this
area to integrate network virtual functions in multi-access edge computing
platforms.
142 Edge Computing

6.2.5 GAMING
From youngsters to old age, many people around the world are drowned in video
games. Though there is a negative impact on few games, it is proved that games
providing abundant benefits to old age people in particular of brain functioning.
In addition, it pulls down the negative mental effects and improves memory
power. On the contrary, the games consume high bandwidth and latency sensi­
tivity in a multi-player gaming environment. By placing edge servers close to the
places, backhaul bandwidth and low latency are promising. The lower the latency,
the better the gaming environment is guaranteed. With edge computing, better
user experience is feasible and it reignites the user experience. Edge-hosted
gaming service at an affordable price is another fortuning service offered to the
users than investing a huge amount in PlayStation/Xbox.

6.2.6 CONTENT DELIVERY

Better experience is guaranteed for accessing a web page, video, or music via edge
computing model as the response is ensured in milliseconds rather than depending
upon the response from the cloud data center. The Content Delivery Network (CDN)
market has been controlled/ruled by players, like Akamai and Limelight by significant
construction of expansive networks. Few providers like Netflix/HBO sprawl their
customized micro-cache at edge network with greater reliability and flexibility.

6.2.7 FINANCIAL SECTOR

To enable Smart Banking facility, banking sectors are adopting edge computing fra­
meworks for better consumer experience and service. The service is offered via dif­
ferent multitudes over smartphone, desktop/laptop. In addition, the feature is expanded
to ATMs and kiosks with the strong inbuilt functionality to collect and process data.
With these wider ranges of features, banking sectors attract customers with much more
responsive design. Edge computing supports trading sectors by speeding up the
communication and computation, thus reduces the overall loss. High-speed execution
is viable by placing the edge servers very close to stock exchanges to run high resource
intensive protocols/algorithms. More accurate and appropriate latest information is
facilitated to the providers and consumers with this high benefitted model.

6.2.8 AUGMENTED REALITY

Augmented reality pays recent attention due to its applicability in more common
practical applications. Augmented reality facilitates a real appearance to the users
by overlaying digital devices/elements on top of real-world objects/environments.
Wearable devices such as headsets, glasses, smart watches, jackets, and rings create
virtual effects most widely via smartphones. The most familiar application of
augmented reality is the filter option in Instagram/Snapchat. The working procedure
behind augmented reality necessitates the elements/devices to process visual data
and to map the visual data to real-time objects. Edge framework works in a better
Edge Computing Use Cases and Case Studies 143

lighter manner that allows devices to fuse with augmented reality displays spon­
taneously without any delay. Many retail chains are making use of this technology
for a better shopping experience.

6.2.9 HEALTHCARE SECTOR

Edge computing fascinates health sectors by creating new opportunities to in­
corporate recent IT solutions for promising patient care. Gathering, protecting, and
processing a humongous amount of patient data becomes real with the aid of edge
computing. Dynamic real-time decisions and actions are possible to provide by the
healthcare sectors even to the rural background people as e-services in milliseconds
of time with the deployment of edge. In addition, anywhere, any time access to
information is possible than depending upon one centralized database. Recently,
many medical devices appear and ruling the market with a provision to collect and
process data that plays a major role during treatment or diagnosis. For instance,
simple prediction of heart disease using low-density lipoproteins in edge computing
systems is proposed by Anitha Kumari et al. [3]. Thus, edge computing creates a
huge impact on every person’s daily life to lead a healthy and safe life.
Apart from this, edge computing lays its footprints in sectors/applications like
video monitoring, video conferencing, software-defined networking, connecting
homes and offices, and retail.
The realization of use cases in edge computing is presented in Fig. 6.2 for a clear
understanding.

smart car

Edge Server

Cloud Data Center

Smart Manufacturing
smart Bank

Smart Hospital
Edge Gateway/Server

FIGURE 6.2 Realization of edge use cases.