Performance analysis of fog based smart surveillance system

ABSTRACT

There has been increasing demand of security and safety in public as well as private places
and hence surveillance system using IOT sensors, cameras has become the most important
part in our daily life. This system has to operate all times and thus produces huge amounts of
data. Cloud computing offers storage, processing, and analytical services for handling of such
massive amounts of data. For real time applications like smart surveillance system, increased
latency from centralized Cloud computing is not acceptable. Fog Computing is an extension
of Cloud computing, evolved to minimize latency. A Fog-Based Smart Surveillance System
has been modelled and simulated in two environments as Cloud Only Network and Fog-
Based Cloud Network using iFogsim. Various performance metrics like Application Loop
Delay, Energy Consumption, Execution Cost, and Network Usage has been compared
between Fog-Based Cloud Network and Cloud Only Network. Results showed that Fog-
Based Cloud Network performs better than Cloud Only Network.

Keywords: Fog Computing, Smart Surveillance System, I)T, Application Loop Delay,
Energy Consumption, Execution Cost, Network Usage, iFogsim

D.C.P.E, H.V.P.M., Amravati Page 1

 Performance analysis of fog based smart surveillance system

INTRODUCTION

During the last few years, we have witnessed that the two major key technologies like the
smart phone, cloud computing has revolutionized the way people communicate and interact
each other and with machine, and added tremendous level of comforts in their lives. The
main cause of success goes mainly to the introduction of cloud computing system that aid in
reducing the burden of computing and storage capacity required for end user devices In
parallel, we also have witnessed the development and implementation of the Internet of
Things (IOT) into our daily lives, in areas such as transportation, health care, agriculture,
emergency response, industrial automation, home automation. Cloud computing is the base
stone for offering infrastructure, platform, and software services to develop IOT enabled
systems that ultimately meets the requirement of future IOT applications. However, there still
exists some issues and challenges in traditional centralized cloud paradigm that needs to be
solved by new cloud paradigm. In centralized cloud system, the cloud data centre usually
resides remotely from end user devices which leads to higher latency. Besides, handling
massive amount of raw data from massive amount of IOT devices add more cost due to the
requirement of higher computing and storage capacity of cloud servers and more network
bandwidth requirement to the cloud servers. Fog computing is a promising solution to address
above challenges in the cloud based IOT. Fog Computing is a new paradigm in cloud
computing that does not eliminate the use of centralized cloud services but rather extends the
cloud to the edge of network and brings the cloud resources closer to the IOT devices,
physically and/or computationally by increasing the layers in between cloud and end user
devices.

Fog Computing works as a middle layer between Cloud and end user IOT devices. In fog
computing, the big data generated by different types of IOT devices and applications can be
processed at the network edge instead of sending them all to the centralized cloud
infrastructure thus reducing computation overload of cloud server, saving network bandwidth,
reducing response time or latency and response location awareness . The cloud network in
which all the modules of an application run only on a centralized cloud data centers is termed
as Cloud only network and the cloud network where most of the modules (delay sensitive) of
an application run on edge devices like router, switches, cellular base stations etc., near or
close to IOT devices and some of the modules (delay tolerant) run on a centralized cloud data

D.C.P.E, H.V.P.M., Amravati Page 2

 Performance analysis of fog based smart surveillance system

centre is termed as Fog-Based Cloud Network. Fog Computing is also sometimes termed as
Edge Computing, Mobile Edge Computing, Cloud lets. Fog computing-based architecture
can also be considered as the future of IOT infrastructure. Mostly Sense Process Actuate
Model (SPAM) is commonly used in Fog Computing applications.

In this SPAM model, fog device gets data from sensors and does processing. Fog device
sends results to actuator after processing to make necessary actions. Fog Devices also
forward some data to Cloud for managing storage and processing for long term analytic and
also if fog devices are unable to handle tasks due to limitation in resource. These fog
computing applications are developed as Directed Acyclic Graph (DAG) that contains
different modules following Distributed Data Flow (DDF) model. In DAG, a module takes
input, processes it, and sends to output to another module as input. Smart mobile phones,
cellular base stations, switches, routers, etc., can serve as fog devices. These fog devices are
resource constrained in terms of processing capability, storage capacity, and limited power.

D.C.P.E, H.V.P.M., Amravati Page 3

 Performance analysis of fog based smart surveillance system

RESEARCH ISSUES

From one viewpoint, the improvement of Fog and Edge mists incorporates devoted
offices, working framework, organize and middleware strategies to manufacture and work
such miniaturized scale server farms that host virtualized registering assets. Then again, the
utilization of Fog expects the expansion to current programming models and proposes new
deliberations that will enable engineers to outline new applications that take advantage from
such enormously conveyed frameworks. The utilization of this approach additionally opens
up different difficulties in: security and protection (as a client currently needs to "trust"
each smaller scale server farm they interface with), bolster for asset administration for
versatile clients who exchange session starting with one miniaturized scale server farm then
onto the next, and bolster for "installing" such smaller-scale server farms into gadgets (e.g.
autos, structures, and so forth.). By adding quality of experience (QOE) domain in Fog
computing, service will be improved for video streaming and game services. The technical
parameters are different in every field such as multimedia services contains bit rate of video,
frame rate, video codes etc. and game required high processing power to adding QOE
will provide user stratification about the services. To fulfill the consistently expanding
interest for Cloud Computing assets from rising applications, for example, Internet-of-
Things (IOT), scholastic and industry specialists are currently pushing for going from
substantial unified Cloud Computing foundations to smaller scale server farms situated at
the edge of the system. These smaller scale server farms are frequently more like a client
(geologically and in getting to dormancy) contrasted with the incorporated cloud server
center. The point of using such edge assets is to offload calculation that would have
"customarily" been completed at the cloud server center to an asset that is more like a client
or edge devices. This vision additionally recognizes the variety in arranging dormancy from
an end client to cloud server center. Though the system around a server center is frequently
high limit and speed that close to the client devices may have fluid properties (Regarding
flexibility, transmission capacity, idleness, and so forth). Cloud and network management are
important for service providers to provide QOS to end users, so they can evaluate their
services via subjective feedback of users, sometimes which contains in accurate and
negative responses

D.C.P.E, H.V.P.M., Amravati Page 4

 Performance analysis of fog based smart surveillance system

BACKGROUND

There are several issues associated to the cloud computing, such as higher response
time and greater communication costs. Compared the cloud and fog computing paradigms
and concluded that the fog lowers the processing delay significantly. Fog computing is
regarded as an excellent solution to fulfill the demands for large number of connections and
low latency applications by performing some computation and processing tasks from the
cloud server to the fog node.
Fog computing has exhibited a tremendous potential to minimize the latency and
network usage issues of cloud computing. Several studies have proposed fog-based
architectures in different domains to make their systems more efficient,secure and cost
effective. presented a fog based architecture that provides unified cloud and fog resources for
deploying network functions, such as virtualization, mobile edge computing and IOT services
at extreme edge of telecommunication operator’s network. Proposed a software-defined
networking based fog computing architecture to support mobility.
A key benefit of fog computing is that it helps in reducing the frequent data
transmissions to the cloud. This results in minimizing the network usage required by the
applications interacting with the the cloud that eventually results in enhancing the response
rate. Many studies have proved that the latency in fog is low as compared to the cloud. For
example, the authors demonstrated that the latency in fog is low when compared to the
cloud-based implementation of the systems. In addition, fog computing reduces the network
traffic and supports scalability , and is highly suitable for IOT frameworks. Network usage is
always a vital parameter for real-time applications and is effectively minimized using the fog
computing.

D.C.P.E, H.V.P.M., Amravati Page 5

 Performance analysis of fog based smart surveillance system

PROBLEM STATEMENT
With the complexity of networks and the large number of devices, service providers need
efficient planning tools to design optimal networks. Efficient network planning tools can help
in designing optimal networks guaranteeing better service to the end users. The fundamental
goal of network planning is to implement an optimum network addressing realistic traffic
parameters. Unlike earlier generations, computer based automatic planning tools are now
required for large networks. The network design problem can be solved by using simulation
or solving mathematical models. Simulation is typically used when a mathematical model
cannot be developed for a network design problem.
A basic representation of the network is modeled in a tool where the simulated network is
examined with different inputs and parameters. Unfortunately, most network planning
problems are NP-hard. Network planning is a massive research field with many applications
in IP networks, optical networks, wireless networks, cellular networks, cloud computing, etc.
To the best of our knowledge, no work has considered the planning and design of fog
networks in collaboration with the cloud. It can be recognized that the fog nodes will be the
distributed fog network entities enabling the deployment of fog services. For a normal fog
network, there will be a lot of fog nodes to be installed and to deliver the expected
performance improvements, proper planning of these fog nodes is necessary. Consequently,
many factors such as the number of nodes, locations, node sizing, connectivity between fog
and cloud, traffic distribution and so on need to be considered. Given the large number of fog
nodes in the network combined with a vast number of edge devices over the entirety of the
geographical region of interest, the search space for an optimal solution is huge. Therefore,
sophisticated methods and algorithms are required to help network planners in the decision
making process. The purpose of the fog network is to improve the end user experience by
reducing parameters like delay in the network, the amount of traffic in the cloud and so on.
Optimizing more than one parameters converts the problem into a multi objective
optimization problem resulting in a number of non-dominated solutions. The multi objective
optimization problem can be studied from different viewpoints with different solution goals
while solving it. Hence, it is very hard to find a best compromised solution for all the
objective functions even with a comprehensive knowledge of the problem. There are different
methods to solve multi objective optimization problems. However, it is a very complex task
to choose the best method as different methods have various requirements and can perform

D.C.P.E, H.V.P.M., Amravati Page 6

 Performance analysis of fog based smart surveillance system

differently with different problems. Ideally, for the best convergence towards the Pareto set,
the optimization method that handles the minimum amount of complexity should be used.
But defining the amount of complexity is not straightforward. To mitigate that, a number of
methods can be used simultaneously to understand and solve the problem. Moreover, the
planning problem can be solved with exact and approximate multi objective algorithms, each
having its own advantages and disadvantages. The fog network planning is a complex but
necessary step towards building efficient fog networks. To that end, this thesis addresses the
above mentioned problems and proposes an exact mathematical model with different multi
objective optimization methods for the planning and design of fog networks.

D.C.P.E, H.V.P.M., Amravati Page 7

 Performance analysis of fog based smart surveillance system

RELATED WORK

Fog Computing is still in its infancy stage. Currently there are several research works
underway focused on improving various performance metrics of the Fog-Cloud Network
environment. A discussion about applications, issues, architecture, and protocols of fog
computing have been presented in several performance metrics for evaluating fog computing
networks can be found in the literature of (Mahmoud et al., 2018), (Rahbari & Nickray, 2019),
and (Nakamura, Duolikun, Oma, Enokido, & Takizawa, 2018) . They include power
consumption, latency, response time, energy cost, and network utilization. Fog computing is
at the nascent stage of development and adoption, and various simulation tools are recently is
use in order to test and validate various technical solutions, resource management policies,
empirical analysis and research works in Fog Computing. In (Svorobej et al., 2019), the
author has reviewed about various Fog modelling and simulation tools, their comparative
study, problems posed by fog computing in relation to simulation. Out of various available
simulation tools (Byrne et al., 2017), iFogsim simulator has gained remarkable attention from
many Fog Computing researchers.
.

D.C.P.E, H.V.P.M., Amravati Page 8

 Performance analysis of fog based smart surveillance system

PROPOSED WORK

The Fog-Based smart surveillance system application is designed and simulated in iFogsim to
perform a comparative study on the performance of the Fog-Based Cloud Network and Cloud
Only implementation. For better accuracy in comparison results, 5 different configuration has
been used. The simulation environment in this research work is done in a personal computer
with Intel Core i7- 5600U CPU @2.6GHz, 8 GB RAM, OS as Microsoft Windows 7 32-bit
using iFogsim . Eclipse IDE Version: 2020-03 (4.15.0), Build id: 20200313-1211 is used as
Java IDE for this research project.

A. Devices and Setup

Figure 1 Hierarchical Topology of Fog-Based Surveillance System

Figure shows a general hierarchical structure of the Smart Surveillance System topology
created using the GUI module in iFogsim. The topology consists of the devices as Cloud Data
Center, Proxy Server, Fog Device (Router), Cameras, Sensors, and Pan-Tilt zoom (PTZ)
Actuators.

D.C.P.E, H.V.P.M., Amravati Page 9

 Performance analysis of fog based smart surveillance system

The Figure 2 shows the application model of the Smart Surveillance System. The work flow
loop in this system is as Camera >>Motion Detector>>Object Detector>>Object Tracker&
User Interface >>PTZ Control.

D.C.P.E, H.V.P.M., Amravati Page 10

 Performance analysis of fog based smart surveillance system

EXPERIMENTAL SETUP
We simulated the scenario that involves the high defination smart cameras to capture the
images. The images are subsequently transmitted to the fog node. The fog node processes the
images to detect the status of the slot and displays the pictures. Slots on a smart LED
connected to the fog node using the Wi-Fi connection. A proxy server establishes the
connection between the fog node(s) and the cloud server. To simulate the scenarios, we used
iFogsim, which is a toolkit for the IOT devices. By using the iFogsim, we can evaluate the
latency, and network usage. We created the variables of areas and a number of cameras in
the simulation. We created three areas in our experimental scenario. In each area, initially
two cameras were assigned to capture the images of the area. We created a primary cloud
device to which our fog nodes are connected using the proxy server and routers. It is essential
to mention that for each area, we created one fog node as router. As the cameras are smart
(WiFi enabled) and connected through micro controller, therefore we created the cameras in
the simulation environment and considered them as sensors. We increased the number of
cameras to evaluate the results for different scenarios. For a particular scenario, we increased
the number of cameras to evaluate the effects on the latency, and total network usage in a fog
node.

D.C.P.E, H.V.P.M., Amravati Page 11

 Performance analysis of fog based smart surveillance system

RESULT AND DISSCUSION

The simulation results include the value of performance metrics like Application Loop Delay,
Energy consumption by devices, Execution cost in cloud, and network usage. The plots and
graphs of the various results are shown in following Figure.

A. Application loop delay

It is observed from Figure that the both application loop delays from Source module
motion_detector to destination module object_detector and then to destination module
object_tracker, and from source module object_tracker to Destination module PTZ_Control
are very much lesser in Fog-Based cloud network than that in Cloud Only Network in all 5
configurations. This result can be justified from the fact that in Cloud only network all the
data are sent to the cloud data centre for processing and analyzing while in Fog-Based Cloud
network most of the data are processed and analyzed at the fog devices and some data are
sent to the cloud when the fog devices are unable to handle tasks or job due to its resource
constraints. Minimization in application loop delay helps application more responsive and
work faster

D.C.P.E, H.V.P.M., Amravati Page 12

 Performance analysis of fog based smart surveillance system

B. Energy Consumption

It is observed from Figure that the energy consumed by cloud and total energy consumption
by all devices in a network are lesser in Fog-Based cloud network than that in Cloud Only
Network in all 5 configurations. This result can be justified from the fact that burden of
processing and analyzing the data are more in Cloud only network than that in Fog-Based
Cloud network. Reduction of energy consumption helps to cut down the operating cost of the
IOT applications.

C. Cost of execution

It is observed from Figure that the cost of execution in Fog-Based cloud network is very
much lesser than that in Cloud Only Network in all 5 configurations. This result can be
justified from the fact that burden of processing and analyzing the data in cloud data centre is
reduced which ultimately reduced total energy consumption in Fog-Based Cloud. Reduction
in execution cost helps to minimize budget for the large scale implementation of IOT
applications in diverse sectors.

D.C.P.E, H.V.P.M., Amravati Page 13

 Performance analysis of fog based smart surveillance system

D. Network Usage

It is observed from Figure 8 that network usages in Fog-Based cloud network is very much
lesser than that in Cloud Only Network in all 5 configurations. This result can be justified
from the fact that

D.C.P.E, H.V.P.M., Amravati Page 14

 Performance analysis of fog based smart surveillance system

FUTURE WORK

In recent times, Fog computing has assumed a vital role, especially in time-sensitive
application domains. Due to the increase in the data generating devices, the need for quicker
response has also been intensifyfied. To that end, we proposed a fog-based smart parking
architecture that uses computer vision technique to identify an empty parking slot thus
enabling drivers to find the parking slot in minimum time that eventually minimizes the time
and fuel consumption. The experimental results reveal that the proposed fog-based
architecture not only minimizes the latency but also reduces the network usage as compared
to the cloud.A limitation of the proposed research is the use of cameras for parking space
detection. This may give rise to the privacy issues for the car owners as the images are stored
in the cloud. Considering the fact that most of the storage and processing is handled on the
nearby fog nodes. Though, there is a need to preserve the privacy of data in cloud storage by
applying suitable encryption techniques and can be an important direction for future work.
Moreover, it is also important to mention that the large-scale application of the proposed
frame work with the increased number of parking areas will require balancing the load on fog
nodes to maintain the efficiency. Therefore, in future we intend to investigate the load
balancing issues in fog nodes and devise an effective solution to resolve the issue.

D.C.P.E, H.V.P.M., Amravati Page 15

 Performance analysis of fog based smart surveillance system

CONCLUSION

This research work has designed and simulated a fog based smart surveillance system in
iFogsim and performed a comparative study on the performance of the Cloud only and Fog -
Based Cloud Network approach in different configurations . The results of the simulation are
embraced to determine the performance highlights of the Fog - Based Cloud Network over
Cloud Only Network approach . The results of the experiment showed that application loop
delay is reduced , energy consumption of the device is lower , execution cost is efficient , and
network usage is fairly lesser in Fog - Based Cloud Network approach than that in Cloud
Only Network approach . As a result , real time security surveillance system is improved in
Fog - Based Cloud Network approach . Fog Computing is a new paradigm and an extension
of Cloud Computing . This better performance results justifies the suitability of IoT
applications using Fog - Based Cloud Network approach .

D.C.P.E, H.V.P.M., Amravati Page 16

 Performance analysis of fog based smart surveillance system

REFERENCES

[1] A. Aliedani, S. W. Loke, A. Desai, and P. Desai, “Investigatingvehicle-to-vehicle

communication for cooperative car parking:The copark approach,” in 2016 IEEE
International Smart Cities Conference (ISC2). IEEE, 2016, pp. 1–8.

[2] A. Khanna and R. Anand, “Iot based smart parking system,”in 2016 International
Conference on Internet of Things andApplications (IOTA). IEEE, 2016, pp. 266–270.

[3] C. Tang, X. Wei, C. Zhu, W. Chen, and J. J. Rodrigues,“Towards smart parking based on
fog computing,” IEEE Access,vol. 6, pp. 70 172–70 185, 2018.

[4] S. Ravishankar and N. Theetharappan, “Cloud connected smartcar park,” in 2017

International Conference on I-SMAC (IoT inSocial, Mobile, Analytics and Cloud)(I-SMAC).
IEEE, 2017,pp. 71–74.

[5] M. Mukherjee, L. Shu, and D. Wang, “Survey of fog computing:Fundamental, network

applications, and research challenges,”IEEE Communications Surveys & Tutorials, vol. 20,
no. 3, pp.1826–1857, 2018.

[6] M. Aazam, S. Zeadally, and K. A. Harras, “Fog computingarchitecture, evaluation, and

future research directions,” IEEECommunications Magazine, vol. 56, no. 5, pp. 46–52, 2018.

[7] Y.-C. P. Chang, S. Chen, T.-J. Wang, and Y. Lee, “Fog comput-ing node system
software architecture and potential applicationsfor nb-iot industry,” in 2016 International
Computer Symposium(ICS). IEEE, 2016, pp. 727–730.

[8] A. Alrawais, A. Alhothaily, C. Hu, and X. Cheng, “Fog comput-ing for the internet of
things: Security and privacy issues,” IEEEInternet Computing, vol. 21, no. 2, pp. 34–42,
2017.

D.C.P.E, H.V.P.M., Amravati Page 17