Fundamentals of Fog and

Edge Computing

BCSE313L
Dr. Nivethitha V
SCOPE
Course Objectives

1. To introduce IoT enabling technologies and its opportunities.

2. To review underlying technologies, limitations, and challenges along

with performance metrics and discuss generic conceptual framework
in fog computing.

3. To impart the knowledge to log the sensor data and to perform

further data analytics.
Course Outcomes
At the end of this course, student will be able to:

1. Explore technologies behind the communication and management of fogs and edge resources.

2. Learn the techniques for storage and computation in fogs, edges, 5G and clouds.

3. Implement Internet of Everything (IoE) applications through fog computing architecture and
use optimization techniques for the same.

4. Analyze the performance and issues of the applications developed using fog and edge
architecture.
 • The number of IOT devices has reached a 8 billion by 2020 which in
turn is generating massive amount of data

• In 2023, 5G has made up to around one-fifth of all mobile data traffic,

with subscription uptake forecast to reach 1 billion.[1]

Introduction • According to the latest estimates Cisco IBSG , we could see more
than 75.44 billion IoT devices by 2025 — a 5X increase since 2015
[2]

• In today’s prevailing IoT architecture, sensor data is transmitted over

a wide area network to be centralized, processed and analyzed
.
• The data consumers also consume the data from the cloud through the
core network.
 Figure 1: Cloud computing paradigm

• In the cloud computing paradigm, the end devices at the edge usually
Burden on the play as data consumer.

Cloud • Data producers generate raw data and transfer it to cloud, and data
consumers send request for consuming data to cloud.

• With the growing quantity of data generated at the edge, speed of data
transportation is becoming the bottleneck for the cloud-based computing
paradigm.

• Unfortunately, this approach is unsustainable in the long term, due to the

number of device connections, volume of data, latency across different
locations and networks, and the asynchronous nature of many
connections between data flow and analytical cloud services.
 Figure 2 : Drawbacks of cloud computing

The massive increase in the data generated and transferred by the Iot brings several
Drawbacks of issues which could degrade the Quality of Service with delays or even failed
requests due to bandwidth limitations.
cloud computing
Data Handling
by the Edge Figure 3: Edge computing paradigm

In the edge computing

Fig. 2 illustrates the two-
paradigm, the things not
way computing streams in
only are data consumers, but
edge computing.
also play as data producers.

Edge can perform

At the edge, the things can computing offloading, data
not only request service and storage, caching and
content from the cloud but processing, as well as
also perform the computing distribute request and
tasks from the cloud. delivery service from cloud
to user.
Leading Tech Companies
Innovating in Fog and
Edge Computing

•Cisco: Developing fog computing

solutions to extend cloud computing to the
edge of the network.
•Amazon Web Services (AWS): Offering
edge computing services like AWS
Greengrass and AWS IoT.
•Microsoft: Providing Azure IoT Edge for
deploying cloud workloads to edge devices.
•Google: Offering Google Cloud IoT Edge
for intelligent edge computing solutions.
Leading Tech Companies
Innovating in Fog and
Edge Computing

•IBM: Implementing edge computing through IBM

Edge Application Manager and IBM Watson IoT.

•Intel: Investing in edge computing technologies

and platforms to enhance data processing at the
network's edge.

•Dell Technologies: Offering edge computing

solutions through its Dell Edge Gateway and
other infrastructure products.

•NVIDIA: Providing edge AI computing platforms

like NVIDIA Jetson for AI and deep learning at the
edge.
Relevant Technologies

Edge Computing
This computing paradigm performs
computation done at the edge of the network
that are close to users

Mist
This computing paradigm describes dispersed
computing at the extreme edge (the IoT devices
themselves) and has been proposed with future
self-aware and autonomic systems in mind.

Fog Computing
Fog computing is a decentralized computing
infrastructure in which data, computing,
storage, and applications are distributed in the
most efficient or logical place between the data
source and the cloud

MACC
Mobile devices in a mobile ad hoc cloud network
are also responsible for routing traffic among
themselves, because of the lack of network
infrastructure. By pooling local mobile resources
MEC
MEC as a platform that provides IT and cloud-
computing capabilities within the Radio Access
Network (RAN) in 4G and 5G, in close proximity
to mobile subscribers
MC
Mobile computing, or nomadic computing, is
when computing is performed via mobile, portable
devices, such as laptops, tablets, or mobile phones.
MCC
MCC is defined as an infrastructure where both
the data storage and data processing occur outside
of the mobile device, bringing mobile computing
applications to not just smartphone users but a
much broader range of mobile subscribers
Cloudlet
Cloudlets are small data centers (miniature clouds)
that are typically one hop away from mobile
devices. The idea is to offload computation from
mobile devices to VM-based cloudlets located on
the network edge
Smart System with Different Computing Architectures
Edge Computing Architecture

■ It distributes certain tasks to the IoT devices or the co-located computers within the
same subnet of the IoT devices

■ Tasks: data classification, filtering, or signal converting etc

Fog Computing Architecture

■ Covers a broad range of equipment and networks

■ Hierarchical-based distributed computing model

■ Horizontal scalability of the computational resources

■ It can distribute simple data classification tasks to IoT devices

■ Assign the more complicated context reasoning tasks at the edge gateway devices

■ The analytics tasks that involve terabytes of data, which require higher processing
power to ??????
Fog Computing Architecture
■ Assign the tasks among the resources across different tiers depends on efficiency and
adaptability
■ Timely notification about the situation – Where to assign?
– Ex: patient’s condition in the smart healthcare
– security state of the smart home
– traffic condition of the smart city
– water supply condition of smart farming
– production line operation condition of a smart factory
FEC COMPLETING
THE CLOUD
 Advantages of FEC : SCALE

Security Cognition Agility

Remote wireless source code Shifts the role of IoT devices IT brings the opportunity to
(security software) update from passive to active smart individual and small
affected by signal strength, devices businesses to participate in
interruptions, constraint Mechanism: self-adaptation, providing FEC services using
bandwidth etc self-organization, self-healing, the common open software
Form the best route among self-expression etc interfaces or open Software
the network via FEC nodes Development Kits (SDKs)
Rapid update Ex: MEC standard of ETSI and
the Indie Fog business model
Advantages of FEC : SCALE

Latency Efficiency
Rapid response Performance
Time sensitive function Reduce the unnecessary cost
Ubiquitous healthcare or elder care: Perform
data analytics at the gateway
It reduces bandwidth utilization cost
How to achieve the advantages: SCANC

■ Storage
■ Compute
■ Acceleration
■ Networking
■ Control
■ Storage
– Temporary unit & cache at FEC nodes
– Ex: Vehicle can utilize roadside FEC nodes to share the fetched date
■ Compute
– 2 Models
– I/Paas: infrastructure or platform as a service
■ VM (or) CE (Containers Engine)
■ Deploy the customized software they need in a sandbox environment
– SaaS: software as a service
■ On-demand Data Processing (ODP): Pre-installed methods
■ Context as a Service (CaaS): Customized data provision – Processed data
delivers information to client
■ Acceleration
– Networking -internet service providers (ISPs), control how data travels
(routing paths) through their network. Customers don't usually get to choose
or customize these paths- FEC nodes (specialized network devices) to
manage multiple routing paths simultaneously
– Computing (GPU, FPGA)
■ Network
-Vertical -FEC nodes enable the vertical network using IP network-based standard
protocols
• FEC nodes, which act as the proxy of cloud, to collect data from them and then
forward the data to the cloud
• operate as the message broker of publish-subscribe-based protocol
--Horizontal - utilizing Bluetooth, ZigBee (based on IEEE 802.15.4), and Z-
Wave on the IoT devices and connecting them to an IP network gateway toward
enabling the connectivity between the devices and the backend cloud.
• cloud can request -FEC server hosted in a connected car communicate with
the roadside IoT equipment using ZigBee in order to collect the
environmental information needed for analyzing the real-time
❑ Control
■ Deployment control
✓ - allows clients to perform customizable software program deployment
dynamically.
✓ move their program from one FEC node to another.
■ Actuation control
✓ hardware specification and the connectivities between the FEC nodes and the
connected devices
✓ delegate certain decisions to FEC nodes
❑ Mediation control
✓ corresponds to the capability of FEC in interacting with external entities owned
by different parties.
❑ Security control
✓ allows clients to control the authentication, authorization, identity, and
protection
HIERARCHY OF FEC
Inner Edge Middle Edge Outer Edge
Countrywide, statewide or LAN Constraint Devices
regional WAN of enterprises, • Gateway device with Integrated Devices
ISPs, Data Centers of EPC virtualization tech to support
(Evolved Packet Core) and FEC IP Gateway Devices
MAN • Local cloud/ local data
Improve QoE: Geo-distributed center / cloudlet
cache and processing Cellular Network
Ex: Google Edge Network of • Cellular network-based FEC
Google Cloud Service will be available in a broad
range of related equipment,
collaborates with ISPs – to from macrocell and microcell
distribute data centers at the BTSs to the indoor cellular
ISPs data centers extension equipment such
ISPs offer local cloud within as picocell and femtocell
the Country base stations
BUSINESS MODEL
X as Service
■ X as Service
✓ infrastructure, platform, software, networking, cache
or storage, and many other types
✓ clients to pay to use the hardware equipment that
supports SCANC
✓ individuals and small businesses can also provide
XaaS
Example
✓ IndieFog-popular consumer as provider (CaP)
service provisioning model in multiple domains
✓ MQL5 Cloud Network distributed computing project
(cloud.mql5.com)- utilizes customer-premises
equipment (CPE) to perform various distributed
computing tasks
Support Service
■ Support Service
➢ software management support
➢ the clients who own the hardware equipment can
pay the support service provider
➢ software installation, configuration, and updates on
the clients’ equipment based on their requirements.
➢ support services for maintenance and technical
support.
Example
■ Cisco provides the fog computing solution, the
clients purchase Cisco’s IOX-enabled equipment
and then pay an additional service fee for software
update and technical support for configuring their
FEC environments
 Application Service
Application Service
✓ application solutions to help their clients in processing
the data within or outside of the client’s operation
environments
✓ application serviceprovider can provide applications
configured across all the involved entities at the edge
networks to provide the analysis in an ultra-low latency
manner.
Example
■ assist local government in real-time traffic control
systems that facilitate the self-driving, connected
vehicles.
■ IndieFog provider -installed Apache Edgent can offer
the built-in stream data classification
OPPORTUNITIES &
CHALLENGES
Out of Box Experience
■ FEC-ready equipment such as routers, switches, IP gateway, or hubs will
be available in the market.
■ Products will feature the out-of-box experience (OOBE) in two forms
• OOBE-based equipment
• OOBE-based software
■ OOBE-Based Equipment
➢ product vendors have integrated the FEC runtime platform with their
products such as routers, switches, or other gateway devices.
➢ can easily configure and deploy FEC applications on the equipment
via certain user interfaces
Out of Box Experience
■ OOBE-Based Software
➢ FEC-compatible devices can purchase and install OOBE-based FEC
software on their devices toward enabling FEC runtime environment.

➢ Microsoft Windows – example

➢ users may have devices in heterogeneous specification and

processing units (e.g. x86, ARM etc.) in which the vendor may need
to provide a version for each type of hardware
Open Platforms
■ OpenStack++

■ WSO2–IoT Server

■ Apache Edgent
System Management
■ Three basic life cycle phases: design, implementation, and adjustment.
■ Design
system administration team –
➢ Identify the ideal location among the three edge tiers for placing
FEC servers.
➢ Describes what kind of resources the FEC servers need
➢ Explains how these FEC servers will interact with the rest of the
system.
➢ Implementation
 System Management
■ Implementation

■ Variety in FEC Environments: The administration team must account for the differences in
the environments where FEC servers are used, particularly in middle-edge and outer-edge areas

■ Different Vendor Platforms: FEC equipment from different vendors (like Cisco or Dell) often
come with their own unique platforms.

■ Open Platform Projects: While there are ongoing efforts to create open platform solutions for
FEC, each platform still has its own specific requirements.
System Management
■ Adjustment
➢ Runtime Adjustment: The FEC system needs to be able to adjust on the fly, deciding
when and where to run its functions to optimize performance.

➢ Dynamic Deployment: The system should be capable of deploying or terminating

runtime environments (like virtual machines or containers) and application methods
on suitable FEC nodes as needed.

➢ Migration of Runtime Environments: The system should also be able to move these
runtime environments or application methods from one FEC node to another based
on real-time factors
06
Edge Computing Platforms:
• Purpose: CORD (Central Office Re-
architected as a Datacenter) aims to
transform network infrastructure into
agile, scalable edge cloud services using
SDN, NFV, and cloud technologies.
• Architecture: It uses commodity servers
and white-box switches in a Spine-Leaf
topology to provide computing, storage,
and networking resources.
• Software Components: Includes
OpenStack for IaaS, Docker for
containerization, ONOS for network
management, and XOS for service
control.
• Deployment: CORD can be deployed in
three solutions: M-CORD (mobile
networks), R-CORD (residential
networks), and E-CORD (enterprise
networks), each tailored for different use
cases
 •Interoperability Framework: Designed for IoT edge
EdgeX Foundry computing, connecting various sensors and devices via
different protocols.
•Hardware Agnostic: Can run on various hardware,
operating systems, and environments, including natively
or in Docker containers.
•Microservices Architecture: Consists of four service
layers (Device Services, Core Services, Supporting
Services, Export Services) and two system services
(System Management, Security)
Apache Edgent

• Programming Model: Apache Edgent is an open-source programming

model for lightweight runtime data analytics, designed for small devices
like routers and gateways at the edge.
• Data Analytics: It focuses on data analytics at the edge, providing APIs
to build edge applications that can filter, split, transform, and process
data streams.
• Deployment: Edgent applications can be deployed on Java 8, Java 7,
and Android environments
1. Device Services Layer: Converts data formats from IoT devices and
sends them to the core services
2. Core Data: Acts as a persistence repository and management service,
storing and managing data collected from IoT devices
3. Command: Provides APIs for command requests from applications to
Device Services
4. Metadata: Manages metadata about IoT objects, including Device
Profiles
5. Registry & Configuration: Centralizes management of configuration and
operating parameters for other microservices
1. Supporting Services Layer: Provides edge analytics and intelligence,
including rules engine, alerting, and scheduling.
• Export Services Layer: Connects EdgeX with external systems,
distributing data to registered clients.
• System Management and Security: Handles installation, upgrades,
monitoring, and data security
 Azure IoT Edge
• Components: Azure IoT Edge consists of
three main components: IoT Edge
modules, IoT Edge runtime, and a cloud-
based interface.
• Functionality: It moves cloud analytics to
edge devices, enabling lower latency and
the use of Azure services like Azure
Functions, Azure Machine Learning, and
Azure Stream Analytics.
• Deployment: Modules are containerized
instances running customer code or Azure
services, managed by the IoT Edge
runtime
• IoT Edge Modules: These are containerized instances running customer
code or Azure services. They execute specific applications and can be
dynamically deployed.
• IoT Edge Runtime: This manages the IoT Edge modules. It consists of two
parts: IoT Edge hub (a local proxy for IoT Hub) and IoT Edge agent (deploys
and monitors modules).
• Cloud-Based Interface: This interface is used to monitor and manage the IoT
Edge modules and runtime from the cloud, allowing for large-scale
deployment and monitoring of edge applications
ParaDrop • Edge Computing Framework: ParaDrop
upgrades existing access points to edge
computing systems, supporting
applications and services like a normal
server.
• Multi-Tenancy and Resource
Utilization: Uses container
virtualization to manage resources
efficiently and isolate applications.
• Dynamic Application Management:
Provides APIs for developers to monitor
and manage system resources, and a
web UI for user interaction
 Purpose: Cachier is an edge caching
Cachier •

system designed for image

recognition applications, aiming to
reduce response time by leveraging
edge nodes for processing.
• Components: It includes the Recognizer
Module for analyzing and matching
images, the Optimizer Module for
predicting request distribution, and
the Offline Analysis Module for
calculating cache delay and accuracy.
• Functionality: Cachier dynamically
adjusts the size of the dataset on edge
nodes based on environment factors to
ensure optimal response time