Cloud Computing - Practical

EL
PT Dr. Sudip Misra
N Associate Professor
Department of Computer Science and Technology
IIT Kharagpur
Email: [email protected]
Website: http://cse.iitkgp.ac.in/~smisra/

Introduction to Internet of Things 1
Contents
 Introduction to Openstack
 Components
 Installation

EL
 Creating a key‐pair and manage security group

PT
 Launce Instances
 Creating an image N
 Accessing and Communicating with instances

Introduction to Internet of Things 2
Introduction to Openstack
 A software to create a cloud unfrastructure
 Launched as a joint project of Rackspace Hosting and NASA in 2010
 Opensource

EL
 Presently many companies are contributing to openstack
 Eg. IBM, CISCO, HP, Dell, Vmware, Redhat, suse, Rackspace hosting

PT
 It has a very large community
 Can be used to develop private cloud or public cloud
N
 Versions:
 Austin, Bexar, Cactus, Diablo, Essex, Folsom, Grizzly, Havana, Icehouse, Juno, 
Kilo, Liberty, Mitaka, Newton, Ocata (Latest)

Introduction to Internet of Things 3
Components
Horizon 
Dashboard

EL
Nova Glance Swift
PT
Neutron Cinder Heat Ceilometer Keystone
N

Introduction to Internet of Things 4
Components contd.
 Keystone
 Identity service
 Provides authentication and authorization

EL
PT
 Horizon
 Dashboard N
 GUI of the software
 Provides overview of the other components

Introduction to Internet of Things 5
Components contd.
 Nova
 Compute service
 Where you launce your instances

EL
PT
 Glance
 Image service N
 Discovering, registering, retrieving the VM
 Snapshots

Introduction to Internet of Things 6
Components contd.
 Swift
 Object storage
 Helps in storing data safely, cheaply and efficiently

EL
PT
 Neutron
 Provides networking service
N
 Enables the other services to communicate with each other
 Make your own network

Introduction to Internet of Things 7
Components contd.
 Cinder
 Block storage
 Virtualizes the management of block service

EL
 Heat

PT
 Orchestration
 Ceilometer N
 Billing
 What service you are using 
 How long are you using

Introduction to Internet of Things 8
Installation
 Can be installed manually or using scripts like Devstack
 We will use devstack

EL
 Steps: 
 Install git ( sudo apt‐get install git )

PT
 Clone devstack ( git clone https://git.openstack.org/openstack‐
N
dev/devstack )
 Go to devstack directory ( cd devstack )

Introduction to Internet of Things 9
Installation contd.
 Open local.conf file and paste the following and save the file
ADMIN_PASSWORD=<YOUR PASSWORD>
DATABASE_PASSWORD =<YOUR PASSWORD>

EL
RABBIT_PASSWORD =<YOUR PASSWORD>
SERVICE_PASSWORD =<YOUR PASSWORD>

PT
HOST_IP=<the IP of your PC>
N
 Run the stack.sh file ( ./stack.sh)
 For uninstallation, go to devstack directory and run unstack.sh file

Introduction to Internet of Things 10
References
 1. https://www.openstack.org/
 https://docs.openstack.org/developer/devstack/

EL
PT
N

Introduction to Internet of Things 11
 EL
PT
N

Introduction to Internet of Things 12
 EL
Sensor-Cloud-Part I
Sensor-as-a-Service

PT
Dr. Sudip Misra
Associate Professor
N
Department of Computer Science and Engineering
IIT Kharagpur
Email: [email protected]
Website: http://cse.iitkgp.ac.in/~smisra/

Introduction to Internet of Things 1
Introduction

 It is not mere integration of sensors and cloud computing
 It is not only “dumping the sensor data into cloud”

EL
PT Cloud
N
Cloud

Introduction to Internet of Things 2
Wireless Sensor Networks (WSNs): Recap

 Contain sensor nodes which sense some physical phenomena from the 
environment
 Transmit the sensed data (through wireless communication)  to a centralized 

EL
unit, commonly known as Sink node
 The communication between Sink node and other sensor nodes in the 

PT
network may be single/multi‐hop
 Sink node further process data 
N

Introduction to Internet of Things 3
Wireless Sensor Networks (WSNs): Recap
Sink Applications
 Target Tracking
Sensing unit  Wildlife Monitoring
 Healthcare

EL
Processing unit  Industrial Applications
Communication unit  Smart Home

PT
 Smart City
Major Components of a   Agriculture
Sensor Node
N  …

Wireless Sensor Networks

Introduction to Internet of Things 4
Cloud Computing: Recap

 An architecture which provides on‐demand computing resources
 Advantages
 Elasticity: Scaling up/down 

EL
 Pay‐per‐use: Payment for the resource as per requirement
 Self Service: Resource can be accessed by self

PT
N

Introduction to Internet of Things 5
Cloud Computing: Services
Cloud‐Clients
App, Web browser, terminal

Software‐as‐a‐Service (SaaS)

EL
Platform‐as‐a‐Service (PaaS)

PT
N
Infrastructure‐as‐a‐Service (IaaS)

Introduction to Internet of Things 6
Cloud Computing: Services

 Software‐as‐a‐Service (SaaS)
 A third party provides a host application over internet
 Example: Microsoft Office 365

EL
 Platform‐as‐a‐Service (PaaS)
 Provide a platform to develop and run applications

PT
 Example: Windows Azure
 Infrastructure‐as‐a‐Service (IaaS)
N
 Provide computing resources
 Example: Storage space

Introduction to Internet of Things 7
Virtualization Concept

 One computer host appears as many computers‐concept of Virtual Machine 
(VM)
 Improve IT throughput and costs by using physical resources as a pool from 

EL
which virtual resources can be allocated.
 Benefit

PT
 Sharing of resources: Same resource can be shared, in turn cost reduction
 Encapsulation: A complete computing environment
N
 Independence: Runs independently of underlying hardware
 Portability: VM Migration

Introduction to Internet of Things 8
Limitations of WSNs
Change of Requirement
 Procurement An example

 Price
 Right vendor Today Tomorrow

EL
 Types of sensor integrated with it
 Deployment

PT
 Right way of deployment
 Right place of deployment Smart Home
N Agriculture
 Maintenance
Result: Change in Sensor type, deployment 
 Post deployment maintenance area, topology design, and many more….
 Battery lifetime

Introduction to Internet of Things 9
Sensor-Cloud: Introduction

 Not only the mere integration of cloud computing and sensor networks, but 
sensor‐cloud is more than that
 Concept of virtualization of sensor node

EL
 Pay‐per‐use
 One sensor node/network appears as many

PT
 A stratum between sensor nodes and end‐users
N

Introduction to Internet of Things 10
 Difference with WSN
Multiple applications/
WSN user users

Sensor‐cloud 
Aggregated data Virtualization

EL
infrastructure

Dedicated to a 
PT Serves multiple 
single user
N
applications

WSN Sensor‐Cloud
Source: S. Misra; S. Chatterjee; M. S. Obaidat, "On Theoretical Modeling of Sensor Cloud: A Paradigm Shift From Wireless Sensor Network," in IEEE
Systems Journal , vol.PP, no.99, pp.1-10

Introduction to Internet of Things 11
 Difference with WSN (Contd.)
Actors and Roles
Attributes WSN Sensor Cloud
Ownership WSN‐user Sensor‐owner

EL
Deployment WSN‐user Sensor‐owner
Redeployment WSN‐user SCSP

PT
Maintenances N WSN‐user SCSP
Overhead WSN‐user SCSP
Usage WSN‐user End‐user
Source: S. Misra; S. Chatterjee; M. S. Obaidat, "On Theoretical Modeling of Sensor Cloud: A Paradigm Shift From Wireless Sensor Network," in IEEE
Systems Journal , vol.PP, no.99, pp.1-10

Introduction to Internet of Things 12
Actors in Sensor-cloud

 End‐users
 Enjoy Se‐aaS through applications as per the requirements.
 Unknown about what and which physical sensor is/are allocated to serve the

EL
application
 Sensor‐owner

PT
 Plays a role from business perspective.
 They purchase physical sensor devices, deployed over different geographical
N
locations, and lend these devices to the sensor‐cloud
 Sensor‐Cloud Service Provider (SCSP)
 A business actor.
 SCSP charges price from the end‐users as per their usage of Se‐aaS.

Introduction to Internet of Things 13
Sensor-cloud: Architecture

 End‐users: Registered themselves, selects 
templates, and request for application(s)
 Sensor‐owner: Deploy heterogeneous/ 

EL
homogeneous physical sensor nodes over 
different geographical location

PT
 SCSP: Plays managerial role
N

Introduction to Internet of Things 14
 Sensor-cloud: View
User Login
Xml 
interpretation

specification
Interaction with 
Application 1

Template 
Dynamic Scaling physical sensor
On‐demand physical 
Data feed Template sensor scheduling

Web portal
display

EL
Vast data storage and  Heterogeneous 
User  Browser  specialized processing pool of physical 
organization Interface sensors
Sensed Energy 

Sensed data
PT
Data feed management, QoS
information
Application 
Application 2 specific real‐time 
On‐demand 
sensor data
N data aggregation

Template specification
User organization view Real View
Source: S. Misra; S. Chatterjee; M. S. Obaidat, "On Theoretical Modeling of Sensor Cloud: A Paradigm Shift From Wireless Sensor Network," in IEEE
Systems Journal , vol.PP, no.99, pp.1-10

Introduction to Internet of Things 15
Work Flow of Sensor-Cloud Source: S. Misra; S. Chatterjee; M. S.
Obaidat, "On Theoretical Modeling of
Sensor Cloud: A Paradigm Shift From
User  SensorML Virtual Sensor  Virtual Sensor  Resource  Wireless Sensor Network," in IEEE
organization interpretor Manager Controller Manger Systems Journal , vol.PP, no.99, pp.1-10

Operations
request Create virtual 
sensor instance Manage
operations
Response

EL
Response
Data request
XML template Decode
Compatible 

PT
sensor  Sensor Physical sensor definition,
scheduling,  resource pool  Virtual sensor
allocation, (WSN) Group definition
N deallocation Client information 
Data Data retrieval Metadata
provisioning Data
Templates
aggregation

Delete Release
virtual sensor instance Release
resource

Introduction to Internet of Things 16
 Case Study: Target Tracking

“We consider a WSN‐based target tracking application, in which a WSN owner

refuses to share the sensed information with an external body, even in
exchange of money. Consequently, any organization that wishes to detect

EL
intrusion within a particular zone has to deploy its own WSN. This leads to a
long‐term investment due to costly network setup and maintenance

PT
overheads. However, in a sensor‐cloud environment, the same organization
can use the same tracking application and still get the service without actually
N
owning the WSN”

Source: S. Misra; S. Chatterjee; M. S. Obaidat, "On Theoretical Modeling of Sensor Cloud: A Paradigm Shift From Wireless Sensor Network," in IEEE
Systems Journal , vol.PP, no.99, pp.1-10

Introduction to Internet of Things 17
 EL
PT
N

Introduction to Internet of Things 18
 EL
Sensor-Cloud-Part II
Sensor-as-a-Service

PT
Dr. Sudip Misra
Associate Professor
N
Department of Computer Science and Engineering
IIT Kharagpur
Email: [email protected]
Website: http://cse.iitkgp.ac.in/~smisra/

Introduction to Internet of Things 1

Management Issues in Sensor-Cloud

 Optimal Composition of virtual sensor nodes
 Data Caching
 Optimal Pricing

EL
PT
N

Introduction to Internet of Things 2

 EL
Optimal Composition of Virtual Sensor

PT
N
Source: S. Chatterjee and S. Misra, “Dynamic Optimal Composition of a Virtual Sensor for Efficient Virtualization Within Sensor‐cloud”,
IEEE ICC 2015. 

Introduction to Internet of Things 3

Introduction

 Efficient virtualization of the physical sensor nodes 
 An optimal composition of VSs
 Consider same geographic region: CoV‐I

EL
 Spanning across multiple regions: CoV‐II

PT
N
Source: S. Chatterjee and S. Misra, “Dynamic Optimal Composition of a Virtual Sensor for Efficient Virtualization Within Sensor‐cloud”,
IEEE ICC 2015. 

Introduction to Internet of Things 4

Why Composition of Virtual Sensor?

 Resource‐constrained sensor nodes

 Dynamic change in sensor conditions
 The composition of virtual sensors are non‐traditional

EL
PT
N
Source: S. Chatterjee and S. Misra, “Dynamic Optimal Composition of a Virtual Sensor for Efficient Virtualization Within Sensor‐cloud”,
IEEE ICC 2015. 

Introduction to Internet of Things 5

CoV-I: Formation of Virtual Sensor

 Optimal formation of Virtual

Sensor (VS) VS
 Homogeneous sensor nodes

EL
within same geographical
boundary

PT
N
Source: S. Chatterjee and S. Misra, "Optimal composition of a virtual sensor 
for efficient virtualization within sensor‐cloud," 2015 IEEE International 
Conference on Communications (ICC), London, 2015, pp. 448‐453

Introduction to Internet of Things 6

CoV-II: Formation of Virtual Sensor Group
VSG
 Formation of Virtual Sensor
Group (VSG)
 Heterogeneous physical sensor VS1

EL
VS2 VS3
nodes across different
geographical locations

PT
N
Source: S. Chatterjee and S. Misra, "Optimal composition of a virtual sensor 
for efficient virtualization within sensor‐cloud," 2015 IEEE International 
Conference on Communications (ICC), London, 2015, pp. 448‐453

Introduction to Internet of Things 7

Performance

EL
PT
N
Source: S. Chatterjee and S. Misra, "Optimal composition of a virtual sensor for efficient virtualization within sensor‐cloud," 2015 IEEE International Conference on 
Communications (ICC), London, 2015, pp. 448‐453

Introduction to Internet of Things 8

 EL
Dynamic and Adaptive Data Caching Mechanism

PT
N
Source: S. Chatterjee, S. Misra, “Dynamic and Adaptive Data Caching Mechanism for Virtualization within Sensor‐Cloud”, IEEE ANTS 2014. 

Introduction to Internet of Things 9

Introduction

 Introduces internal and external caching mechanisms 
 Ensures efficiency in resource utilization
 Flexible with the varied rate of change of the physical environment

EL
PT
N
Source: S. Chatterjee, S. Misra, “Dynamic and Adaptive Data Caching Mechanism for Virtualization within Sensor‐Cloud”, IEEE ANTS 2014. 

Introduction to Internet of Things 10

Why Caching in Sensor-Cloud?

 End‐users request for the sensed information through a Web‐interface

 Allocation of physical sensor nodes and virtualization takes place
 Physical sensor nodes continuously sense and transmit data to sensor‐cloud

EL
PT
N
Source: S. Chatterjee, S. Misra, “Dynamic and Adaptive Data Caching Mechanism for Virtualization within Sensor‐Cloud”, IEEE ANTS 2014. 

Introduction to Internet of Things 11

Why Caching in Sensor-Cloud? (Contd.)

 Practically, in some cases, the change in environmental condition are

significantly slow
 Due to the slow change in environment, the sensed data of physical sensors

EL
unaltered
 In such a situation, unnecessary sensing causes energy consumption

PT
N
Source: S. Chatterjee, S. Misra, “Dynamic and Adaptive Data Caching Mechanism for Virtualization within Sensor‐Cloud”, IEEE ANTS 2014. 

Introduction to Internet of Things 12

External and Internal Caching Mechanism

 Internal Cache (IC)

 Handles requests from end‐user
 Takes decision whether the data should be provided directly to the end

EL
user or is it required to re‐cache the data from external cache
 External Cache (EC)

PT
 After every certain interval data are required to re‐cache
 Initially, few data are used to be transmitted to IC
N
Source: S. Chatterjee, S. Misra, “Dynamic and Adaptive Data Caching Mechanism for Virtualization within Sensor‐Cloud”, IEEE ANTS 2014. 

Introduction to Internet of Things 13

Architecture of Caching App1 App2 . . . Appn
App1 App2 . . . Appn

Sensor‐Cloud
Sensor‐Cloud IC

EL
Resource 
EC

PT
pooling
N
Existing Architecture Cache‐enabled Architecture
Source: S. Chatterjee, S. Misra, “Dynamic and Adaptive Data Caching Mechanism for Virtualization within Sensor‐Cloud”, IEEE ANTS 2014. 

Introduction to Internet of Things 14

Performance

EL
PT
N
Source: S. Chatterjee, S. Misra, “Dynamic and Adaptive Data Caching Mechanism for Virtualization within Sensor‐Cloud”, IEEE ANTS 2014. 

Introduction to Internet of Things 15

 EL
Dynamic Optimal Pricing for Sensor-Cloud
Infrastructure
PT
N
Source: S. Chatterjee, R. Ladia, and S. Misra, “Dynamic Optimal Pricing for Heterogeneous Service‐Oriented Architecture of Sensor‐Cloud 
Infrastructure”, IEEE TSC 2017. 

Introduction to Internet of Things 16

Introduction

 Existing schemes consider homogeneity of service (e.g. for IaaS, SaaS)
 No scheme for SeaaS. 
 The proposed pricing scheme comprises of two components:

EL
 Pricing attributed to hardware (pH)
 Pricing attributed to Infrastructure (pI)

PT
 Goal of the proposed pricing scheme:
 Maximizing profit of SCSP
N
 Maximizing profit of sensor owner
 End users’ satisfaction

Introduction to Internet of Things 17

Pricing in Sensor-Cloud Set of end users
Set of sensor owner

Web Portal

EL
PT
Base 
N Station Sensor‐Cloud

Pricing and negotiation

Introduction to Internet of Things 18

Focus on

 Maximizing the profit made by SCSP
 Optimal pricing to the end‐users
 End users satisfaction

EL
 Pricing attributed to hardware (pH)
 Deals with usage of physical sensor nodes

PT
 Pricing attribute to infrastructure (pI)
 Deals with the price associated with infrastructure of sensor‐cloud
N

Introduction to Internet of Things 19

References

 Madoka Yuriyama and Takayuki Kushida , “Sensor‐Cloud Infrastructure ‐ Physical Sensor Management with 

Virtualized Sensors on Cloud Computing”, Research Report , IBM Research ‐ Tokyo IBM Japan, Ltd., 2010 
(http://domino.research.ibm.com/library/cyberdig.nsf/papers/70E4CC6AD71F2418852577670016F2DE/$File
/RT0897.pdf)

EL
 S. Chatterjee, R. Ladia and S. Misra, "Dynamic Optimal Pricing for Heterogeneous Service‐Oriented 
Architecture of Sensor‐Cloud Infrastructure," in IEEE Transactions on Services Computing, vol. 10, no. 2, pp. 

PT
203‐216, 2017
 S. Chatterjee and S. Misra, "Optimal composition of a virtual sensor for efficient virtualization within sensor‐
cloud," 2015 IEEE International Conference on Communications (ICC), London, 2015, pp. 448‐453
N
 S. Misra; S. Chatterjee; M. S. Obaidat, "On Theoretical Modeling of Sensor Cloud: A Paradigm Shift From 
Wireless Sensor Network," in IEEE Systems Journal , vol.PP, no.99, pp.1‐10

Introduction to Internet of Things 20

 EL
PT
N

Introduction to Internet of Things 21

 EL
Fog Computing – Part I

PT
Dr. Sudip Misra
Associate Professor
N
Department of Computer Science and Technology
IIT Kharagpur
Email: [email protected]
Website: http://cse.iitkgp.ac.in/~smisra/

Introduction to Internet of Things 1

Introduction
 Fog computing or fogging is a term coined by CISCO.
 The idea of fog computing is to extend the cloud nearer to the IoT devices.

EL
 The primary aim: solve the problems faced by cloud computing during IoT
data processing.

PT
 an intermediate layer between cloud and devices.
N

Introduction to Internet of Things 2

Introduction (contd.)

Cloud

EL
Fog

PT
N Device

Fig. Fog as intermediate layer between cloud and device 

Introduction to Internet of Things 3

Introduction (contd.)
 40% of the whole worlds data will come from sensors alone by 2020.
 90% of the world’s data were generated only during the period of last two 

EL
years.
 2.5 quintillion bytes of data is generated per day.

PT
 total expenditure on IoT devices will be $1.7 Trillion by 2020
N

Introduction to Internet of Things 4

Introduction (contd.)
 the total number of connected vehicles worldwide will be 250 millions by 
2020.
 there will be more than 30 billion IoT devices

EL
 The amount of data generated by IoT devices is simply huge.

PT
N

Introduction to Internet of Things 5

Why Fog Computing
 The ability of the current cloud model is insufficient to handle the 
requirements of IoT. 
 Issues are:

EL
 Volume

PT
 Latency
 Bandwidth 
N

Introduction to Internet of Things 6

Why Fog Computing (contd.)

Cloud

EL
Sends data  Sends back 

PT
for analysis  command or 
and storage action 
required
N
Devices

Fig.1: Present day cloud model

Introduction to Internet of Things 7

 Why Fog Computing (contd.)
 Data Volume:
 By 2020, about 50 billion devices will be online.

EL
 Presently billions of devices produce exabytes of data everyday.
 Device density is still increasing everyday.

PT
 Current cloud model is unable to process this amount of data.
N

Introduction to Internet of Things 8

Why Fog Computing (contd.)

 Private firms, Factories, 
airplane companies produces  Cloud

EL
colossus amount of data 
Storing data
everyday

PT
 Current cloud model cannot 
store all these data
 Data need to be filtered
N
Private firms Factories
Airplane firms

Introduction to Internet of Things 9

Why Fog Computing (contd.)
 Latency
 Time taken by a data packet for a round trip
 An important aspect for handing a time sensitive data.

EL
 If edge devices send time sensitive data to cloud for analysis and wait 
for the cloud to give a proper action, then it can lead to many 

PT
unwanted results.
 While handling time sensitive data, a millisecond can make a huge 
N
differences.

Introduction to Internet of Things 10

Why Fog Computing (contd.)
Cloud
 Sending time‐sensitive data to cloud for  Analysis of 
data
analysis
 Latency =  + 

EL

+ 
where T = Time

PT
Sending time 
 Latency will be increased sensitive data for  Appropriate 
analysis action
 When the action reaches the device,
N
accident may  have already 
occured

Introduction to Internet of Things 11

Why Fog Computing (contd.)
 Bandwidth:
 Bit‐rate of data during transmission

EL
 If all the data generated by IoT devices are sent to cloud for storage 
and analysis, then, the traffic generated by these devices will be 

PT
simply gigantic.
 consumes almost all the bandwidths.
N
 Handling this kind of traffic will be simply a very hard task.

Introduction to Internet of Things 12

Why Fog Computing (contd.)

 Billions of devices consuming bandwidth Cloud
 If all the devices become online even IPv6 

EL
will not be able to provide facility to all 
the devices

PT
 Data may be confidential which the firms  Sending data for 
analysis and  Appropriate 
do not want to share online
N storage action

Introduction to Internet of Things 13

Requirements of IoT
 Reduce latency of data:
 Appropriate actions at the right time prevents major accidents machine failure 
etc.

EL
 A minute delay while taking a decision makes a huge difference
 Latency can be reduced by analyzing the data close to the data source

PT
N

Introduction to Internet of Things 14

Requirements of IoT (contd.)
 Data security:
 IoT data must be secured and protected from the intruders.

EL
 Data are required to be monitored 24x7
 An appropriate action should be taken before the attack causes major 

PT
harm to the network
N

Introduction to Internet of Things 15

Requirements of IoT (contd.)
 Operation reliability:
 The data generated from IoT devices are used to solve real time 

EL
problem
 Integrity and availability of the data must be guaranteed

PT
 Unavailability and tampering of data can be hazardous
N

Introduction to Internet of Things 16

Requirements of IoT (contd.)
 Processing of data at respective suitable place: 
 Data can be divided into three types based on sensitivity
 time sensitive data

EL
 less time sensitive data
 data which are not time sensitive

PT
 Extremely time sensitive data should be analyzed very near to the data 
source
N
 Data which are not time sensitive will be analyzed in the cloud.

Introduction to Internet of Things 17

Requirements of IoT (contd.)
 Monitor data across large geographical area:
 The location of connected IoT devices can be spread across a large 

EL
geographical region
 E.g. monitoring the railway track of a country or a state

PT
 the devices are exposed to the harsh environments condition
N

Introduction to Internet of Things 18

When should we use fog
 If the data should ne analyze with fraction of second
 If there are huge number of devices 

EL
 If the devices are separated by a large geographical distance
 If the devices are needed to be subjected to extreme conditions

PT
N

Introduction to Internet of Things 19

 EL
PT
N

Introduction to Internet of Things 20

 EL
Fog Computing – Part II

PT
Dr. Sudip Misra
Associate Professor
N
Department of Computer Science and Technology
IIT Kharagpur
Email: [email protected]
Website: http://cse.iitkgp.ac.in/~smisra/

Introduction to Internet of Things 1

Architecture of Fog
 Cloud services are extended to IoT devices through fog
 Fog is a layer between cloud and IoT devices

EL
 many fog nodes can be present
 Sensor data are processed in the fog before it is sent to the cloud

PT
 Reduces latency, save bandwidth and save the storage of the cloud
N

Introduction to Internet of Things 2

Architecture of Fog (contd.)

EL
PT
N

Introduction to Internet of Things 3

Fog nodes
 Characteristics for a fog node:
 Storage ‐ To give transient storage

EL
 Computing facility
‐ To process the data before it is sent to cloud

PT
‐ To take quick decisions
 Network connectivity ‐ To connect with IoT devices, other fog nodes 
N
and cloud

Introduction to Internet of Things 4

Fog nodes (contd.)
 E.g. ‐ routers, embedded servers, switches, video surveillance cameras, 
etc.
 deployable anywhere inside the network.

EL
 Each fog nodes have their aggregate fog node.

PT
N

Introduction to Internet of Things 5

Working of Fog
 Three types of data
 Very time‐sensitive data

EL
 Less time‐sensitive data
 Data which are not time‐sensitive

PT
 Fog nodes works according to the type of data they receive.
 An IoT application should be installed to each fog nodes
N

Introduction to Internet of Things 6

Working of Fog (contd.)
Sends the summary for historical analysis and storage

Cloud
Sends the summary for 
historical analysis and 
Non‐time‐sensitive 

EL
storage
data
Nearest  Sends the summary for 

PT
Fog Node Less time‐sensitive  historical analysis and 
data storage
If time‐sensitive 
Ingest data
N
data then take  Aggregate fog 
immediate action node
Devices
Action
Fig : Working of fog

Introduction to Internet of Things 7

Working of Fog (contd.)
 The nearest fog node ingest the data from the devices.
 Most time‐sensitive data

EL
 Data which should be analyzed within fraction of a second
 Analyze at the nearest node itself

PT
 Sends the decision or action to the devices
 Sends and stores the summary to cloud for future analysis
N

Introduction to Internet of Things 8

Working of Fog (contd.)
 Less time‐sensitive data
 Data which can be analyzed after seconds or minutes

EL
 Are sent to the aggregate node for analysis
 After analysis, the aggregate node send the decision or action to the 

PT
device through the nearest node
 The aggregate node sends the summary to cloud for storage and 
N
future analysis.

Introduction to Internet of Things 9

Working of Fog (contd.)
 Non‐time‐sensitive data
 Data which can be wait for hours, days, weeks

EL
 Sent to cloud for storage and future analysis.
 Those summaries from fog nodes can be considered as less time 

PT
sensitive data.
N

Introduction to Internet of Things 10

Working of Fog (contd.)

Fog node closest to  Fog aggregate nodes Cloud

EL
devices

Analysis duration Fraction of second Seconds to minutes Hours to weeks

PT
IoT data storage Transient Hour, days Months to years
duration
N
Geographical  Very local Wider  Global
coverage

Introduction to Internet of Things 11

Advantages of Fog
 Security
 Provides better security

EL
 Fog nodes can use the same security policy 
 Low operation cost

PT
 Data are processed in the fog nodes before sending to cloud
 Reduces the bandwidth consumption
N

Introduction to Internet of Things 12

Advantages of Fog (contd.)
 Reduces unwanted accidents
 Latency will be reduce during decision making

EL
 Quick decision making
 Better privacy

PT
 Every industry can analyze their data locally
 Store confidential data in their local servers
N
 Send only those data which can be shared to the cloud

Introduction to Internet of Things 13

Advantages of Fog (contd.)
 Business agility
 Fog application can be easily developed according to tools available

EL
 Can be deployed anywhere we need
 Can be programed according to the customer’s need

PT
 Support mobility
 Nodes can be mobile
N
 Nodes can join and leave the network anytime

Introduction to Internet of Things 14

Advantages of Fog (contd.)
 Deployable in remote places
 Can be deployed in remote places

EL
 Can be subjected to harsh environmental conditions
 Under sea, railway tracks, vehicles, factory floor etc

PT
 Better data handling
 Can operate with less bandwidth
N
 Data can be analyzed locally
 Reduce the risk of latency

Introduction to Internet of Things 15

Applications of Fog
 Real time health analysis
 Patients with chronic illness can be monitored in real time

EL
 Stroke patients
 Analyze the data real time

PT
 During emergency, alerts the respective doctors immediately
 Historical data analysis can predict future dangers of the patient
N

Introduction to Internet of Things 16

Applications of Fog (contd.)
 Intelligence power efficient system
 Power efficient

EL
 Reports detail power consumption report everyday
 Suggest economical power usage plan

PT
N

Introduction to Internet of Things 17

Applications of Fog (contd.)
 Real time rail monitoring
 Fog nodes can be deployed to railway tracks

EL
 Real time monitoring of the track conditions
 For high speed train, sending the data in cloud for analysis is inefficient 

PT
 Fog nodes provide fast data analysis
 Improve safety and reliability
N

Introduction to Internet of Things 18

Applications of Fog (contd.)
 Pipeline optimization
 Gas and oils are transported through pipelines

EL
 Real time monitoring of pressure, flow, compressor is necessary
 Terabytes of data are created

PT
 Sending all this data to cloud for analysis and storage is not efficient
 Network latency is not acceptable
N
 Fog is a solution 

Introduction to Internet of Things 19

Applications of Fog (contd.)
 Real time wind mill and turbine analysis
 Wind direction and speed analysis can increase output

EL
 Data can be monitored real time

PT
N

Introduction to Internet of Things 20

Challenges
 Power consumption
 Fog use addition nodes
 Power consumption is higher than centralized cloud

EL
 Data Security
 Data generating nodes are distributed

PT
 Providing authentication and authorization system for the whole nodes is 
not an easy task N
 Reliability
 Maintaining data integrity and availability for millions of nodes is difficult
 failure of a node cannot affect the network

Introduction to Internet of Things 21

Challenges (contd.)
 Fault tolerance
 Failure of a node should be immediately fixed

EL
 Individual failure should not affect the whole scenario 
 Real time analysis

PT
 Real time analysis is a primary requirement for minimizing latency
 Dynamic analysis and decision making reduces danger and increase output
N
 Monitor huge number of nodes is not easy

Introduction to Internet of Things 22

Challenges (contd.)
 Programming architecture
 Fog nodes may be mobile

EL
 Nodes can connect and leave the network when necessary
 Many data processing frameworks are statically configured

PT
 These frameworks cannot provide proper scalability and flexibility
N

Introduction to Internet of Things 23

Conclusion
 Fog is a perfect partner for cloud and IoT
 Solves the primary problems faced by cloud while handling IoT data

EL
 Benefits extends from an individual person to huge firms
 Provides real time analysis and monitoring

PT
N

Introduction to Internet of Things 24

References
 Amir Vahid Dastjerdi, Rajkumar Buyya, Fog Computing: Helping the Internet of Things 
realize its potential, IEEE Fog computing, August 2016
 CISCO white paper, Fog Computing and the Internet of Things: Extend the Cloud to 
Where the Things Are, 2015

EL
 R System white paper, Fog Computing for Big Data analytics, October 2016
 Redowan Mahmud,Rajkumar Buyya, Fog Computing: A Taxonomy, Survey and future 

PT
Directions,Cornell University Library, November 2016
 http://www.businessinsider.in/THE‐INTERNET‐OF‐EVERYTHING‐2015‐SLIDE‐
DECK/articleshow/45695215.cms
N
 http://www.gartner.com/newsroom/id/2970017
 https://www.uschamberfoundation.org/bhq/big‐data‐and‐what‐it‐means
 https://www.afcea.org/content/?q=node/12239

Introduction to Internet of Things 25

 EL
PT
N

Introduction to Internet of Things 26