International Journal of Soft Computing and Artificial Intelligence, ISSN: 2321-404X, Volume-6, Issue-1, May-2018

http://iraj.in
FOG COMPUTING IN IOT
1
TILAK PATEL, 2KRUPA JARIWALA
1
B.TECH STUDENT, S.V.N.I.T, SURAT, GUJARAT, INDIA
2
ASSISTANT PROFESSOR, S.V.N.I.T, SURAT, GUJARAT, INDIA
E-mail: 1tilakp2108@gmail.com, 2 knj@coed.svnit.ac.in

Abstract - Even though the demand of cloud computing is on rise, it involves an inherent problems related to very less
mobility as well as location-awareness, along with unreliable latency. These drawbacks make cloud computing very
unfeasible solution for applications like IoT. It is fog computing which addresses these difficulties as it provides very elastic
resource as well as services till the end devices at the very edge of network, and cloud computing will take more care of
fulfilling demand of resources distributed at the core of the network. Computer scientists and researchers had recently gained
interest in this paradigm of remote computing as it provide computational efficiency of a cloud and bandwidth efficiency of
a local network. This paper reviews the state of the art of FOG computing, its challenges, application and its comparison
with cloud.

Keywords - Internet of Thing, Fog Computing, Cloud Computing

I. INTRODUCTION environment, where they are deployed and use the

results of the analysis to control the environment by
More and more ubiquitously connected smart devices sending commands to environment controlling
are coming in existence and becoming the main factor objects i.e. actuators.
of computing. Various wearable devices, smart
metering machines, smart home, smart city, smart II. FOG COMPUTING
vehicles, large-scale WSN, etc. are coming into
existence. The Internet of Things (IoT) had been a Fog Computing is a distributed platform which can
popular research subject for many years and it will be provide computing, as well as storage and also
considered as the future of Internet. Even after networking services between the IoT devices and
several attempts to augment IoT applications with the conventional Cloud Computing. Data Centers which
power of cloud, problems related to IoT applications either reside at the edge of network or as internal
like low latency, more mobility, geo-distribution, nodes of the distributed environment. Fog Computing
geographic location-awareness etc. are still a major is defined as a scenario where a several ubiquitous
concern[1]. Fog computing can address such and decentralized devices are capable to perform
problems and provides elastic resources and services varied tasks in wireless and sometimes autonomous
to the end users residing at the edge of network so the manner can communicate as well as cooperate in
cloud could focus more about providing resources between them through the network in order to
distributed in the core network. perform tasks of processing and storing data even
Internet of Things (IoT) is a network of without any intervention of third party. These tasks
interconnected objects embedded within everyday can vary from supporting basic network functions or
objects, which are equipped with ubiquitous some new service and applications that run in its
intelligence. There are huge range of devices which virtual environment.
could be wireless sensors and which can just sense
data and provide a continuous stream of small amount
of data or something as complex as smartphones,
tablet, which can produce huge amount of data. These
groups of sensors and along with small, low-power
computational devices are called “things” and
Internet of Things aims to develop a network
infrastructure where these “things” communicate with
each other. It will allow several parameters of
different objects to be collected and several
parameters to be controlled away from environment
through the existing network of infrastructure so we
can quickly integrate the real world along with
computer simulation systems due to which it will
increase efficiency, as well as accuracy and also
economic benefit and reduce the amount of human
intervention. The information gathered by these
Fig 1 Fog Computing [7]
embedded systems are used for analysis of the
Fog Computing in IoT

1
 International Journal of Soft Computing and Artificial Intelligence, ISSN: 2321-404X, Volume-6, Issue-1, May-2018
http://iraj.in
limit. Also, the performance of the client machines
Fog computing is a paradigm of distributed are monitored by the edge device and send graphics
computing which extend its services given by the of appropriate resolution taking care of rendering
cloud datacenters at the IoT devices of the network. time required by the browser.
Fog computing will also give facility like automating 4. Software Defined Networks (SDN): SDN
management of computing, as well as networking and is an important concept based on computing and
storing of data in between the cloud data centers and networking. SDN along with fog computing together
IoT devices. Fog computing consists of several will be capable to resolve some of the main issues
components of such applications which can run both related networks of vehicles, intermittent nature of
on the cloud as well as in edge devices between connectivity, high collision rate and high rate of
sensors and the cloud. Fog computing provides packet losses. Augmenting the vehicle-to vehicle as
features like mobility, computational resources, well as vehicle-to infrastructure communications
network and communication protocols, interface along with main control supported by the SDN does
heterogeneity and connectivity to the cloud as well as this. It would split the control and communication
data analytics of distributed network which takes care layers where controlling will be done by central
of many requirements of many applications having server and server would decide the communication
demands like low latency along with wide as well as path for nodes.
dense geographical distribution. Following are few of
the applications using fog computing. III. ROLE OF FOG COMPUTING IN IOT
1. Healthcare: An analytics system assisted by
fog computing called FAST was proposed in Cao [2] The Internet of Things (IoT) promises to make many
is able to monitor fall condition for stroke patients. items which including devices of consumer
They developed a set of fall detection algorithms, electronics, appliances used in home, devices used in
which uses data like measurements of acceleration medical, different kinds of cameras, and many all
and time series analysis methods along with data types of sensors to be a part of the Internet
noise reducing algorithms to allow increase in environment. This will be opening the door wide to
efficiency in detecting fall condition. It detects fall innovations which will be facilitating many new
condition in real-time and it was based on distributed kinds of interactions among humans and things and it
network of fog computing. will also be enabling us to realize of smart cities, as
2. Augmented Reality: Applications based on well as smart infrastructures and smart services that
augmented reality cannot allow even minor latency as will be enhancing the quality of life. Fog is such an
even very small delays in response can potentially architecture that can distribute computation,
damage the user experience. Fog computing will be communication, control and storage closer to the end
one of the major players of augmented reality domain users along the cloud-to-things continuum.
due to its distributed computing capabilities.
Augmented Brain Computer Interaction Game A. Demands of Internet of Things (IoT)
proposed by Zao [3] is based on Fog Computing and Just like all other technologies, even IoT has its own
Linked Data, it will collect raw streams of data demands. These demands vary for different devices
created by EEG sensors and then it will be classified as there are tons of devices, which are used to create
to detect the brain state of the user while playing a and develop IoT network. Following are the few
game, which uses augmented reality [5]. Brain state demands, which are necessary for any IoT device to
classification is one of the most computationally work efficiently and fulfill our requirements.
heavy signal processing tasks, which needs to be 1. Ubiquitous Devices: Huge demand of IoT
carried out in real-time. devices will increase the number of devices getting
3. Caching and Preprocessing: Zhu[4] connected over the network. The increase of this is
improves website performance by using edge servers mainly due to these two factors: devices used by
where users used to connect with internet by “fog users as well as the micro-sensors and various
boxes” where each of user’s HTTP request goes actuators. Wearable computing devices (smart
through a fog device. The fog device on user’s page watches, glasses, etc.), smart-cities, various
loading requests, to reduce its amount of time, intelligent metering devices used by energy suppliers
performs various optimizations. It will have some to analyze power consumption at an individual house,
general time saving techniques like caching HTML etc. will be the major cause for increase in demand of
components, reorganizing the composition of these devices [1]. SPP-lightweight and efficient
webpages as well as reducing the size of elements in strong privacy preserving authentication scheme for
the web. The edge devices could also perform secure VANET communication.
different optimizations that will analyze the user's 2. Physical Dimension and Power
behavior and internet conditions. For example, when Consumption: Production cost is an important factor
there is congestion in network, the device at the edge due to which device should be as small as possible.
then sends low resolution graphics and photos to the This increases the portability of device and reduces
user so that response time can be under acceptable power consumption, which can be crucial in some

Fog Computing in IoT

2
 International Journal of Soft Computing and Artificial Intelligence, ISSN: 2321-404X, Volume-6, Issue-1, May-2018
http://iraj.in
contexts (for example, cell phones or battery operated other fog devices, IoT and 5G devices, and with end-
fire detectors in a remote forest). Packaging and user input.
energy management technologies aims in creating 4. SCALE: SCALE (Security, Cognition,
smaller and autonomous devices that can run for long Agility, Latency, and Efficiency) is an acronym
time at low price. Even though convenient packages developed by OpenFog. These features are main
improve energy consumption, it may not be enough factors of fog differentiation. IoT differs from others
to last longer. IoT devices require long lasting sensors by resource demands as it has high value applications
that may not be connected to a regular power source. but significant network bandwidth limitations and
Lithium-ion batteries are used today for all kinds of peripherals with limited resources. Interrupted
portable devices. Even so, batteries working on services or intermittent connectivity to the cloud are
electrochemical reactions will become a limiting not tolerable.
factor in future developments. 5. Low Latency : The fog services have small
3. Network management: As all the IoT latency[10] that enables it to perform in real time
devices need to communicate with each other, computations of systems such as AI (Artificial
networking infrastructure between them plays a Intelligence), VR(Virtual Reality), real-time Safety
crucial role. Most of the devices have very less control loops, analytics of streaming of data, etc.
network bandwidth to spare for transmission as well
as to receiving data from other devices. Also, some IV. FOG IMPLEMENTATION IN IOT
devices need a connection with very less latency of
just few milliseconds like actuators to take There are various implementations of fog computing
controlling decisions dynamically. Sensors on the in Internet of Things (IoT) where it had helped in
other hand transmits few kilobytes of data per second, leveraging the benefits of it. Figure 2 shows an
but it is sent continuously round the clock equating it example of fog network which can be implemented to
to several gigabytes of data transfer within a single build a smart city. Each fog node is capable of
day. performing resource intensive task from the data it
gets from its peer nodes as well as from the lower
B. Need for Fog Computing in IoT? level nodes or IoT devices.
Distributed structure of the IoT networks across many
different varieties of industries is very vast in scale
and it will be going to be critical to ongoing
economic growth. Their requirements are operational
efficiencies with zero downtime as well as ultra-low
latency. Also, enormous amount of data loads coming
from billions of streaming devices require processing
power that is closer in proximity to the “things”.
Following are the ways how fog can fill this gap
between cloud and things.
1. Assignment of Functions: Fog computer
effectively distributes the allocation of functions in
IoT networks, unlike other resources. This is able to
simplify as well as standardize the basic operations
related to the global level of IoT network, such as
configuration and administration. Fig. 2 Fog network implementation in a smart city [9]
2. Distributed Architecture: Fog networks
are able to distribute as well as integrate the Figure. 2 shows the fog nodes residing at the
computing, as well as communication, also the endpoints aggregate the data provided by IoT devices
storage and control very flexibly along continuum of before it is sent to the central fog nodes deployed into
cloud for the device. This will also provide a common a building or a street. This helps in minimizing the
methodology for meeting the basic requirements of amount of data transfer and avoids the problems of
IoT, such as computation as well as storage. The network congestion at the upper layers of the
distributed design provides worth across the complete hierarchy. The fog nodes are also responsible to
network not simply on the edge. manage control and scalability of the IoT devices for
3. Immersive distribution: The distributed tasks which require very low latency such as
fog network architecture provides an "immersive biometric identification for which user authentication
distribution", unlike the cloud, which offers need to be done within few seconds or anti-collision
centralization. Immersive distribution means that fog system in trains which need to take decisions in real
resources are available throughout the network. This time.
allows management flexibility and ease of integration Some processes may require lot of computational
with existing IoT environments. Fog can do this by resource which a single fog node may not be able to
interacting not only with the cloud, but also with fulfill alone as well as it cannot be processed onto a
Fog Computing in IoT

3
 International Journal of Soft Computing and Artificial Intelligence, ISSN: 2321-404X, Volume-6, Issue-1, May-2018
http://iraj.in
cloud for e.g. applications related to augmented to have a P2P (peer-to-peer) connection by which it
reality requires huge amount of computational can achieve load balancing, network resilience, fault
resource as well as the results need to be delivered in tolerance, etc.
real time. In such a situation the fog nodes also need

V. COMPARISION OF FOG COMPUTING AND CLOUD COMPUTING OVER IOT

Even though fog computing gives great advantages to the IoT infrastructure, however cloud computing is one of
the emerging solution and it is already in use in many different areas. Also lot of research and development had
already happened into cloud computing compared to fog computing. Fog and cloud are both good solutions but
they are complement of each other in the form of providing service. Table 1 shows comparison of Fog and cloud
computing.

Areas[9] Cloud Fog

Distributed along large
Geo-graphical areas and it is closer to the
Location and model of Centralized in a small
user. Fog nodes and systems can be
Computing number of data centers
controlled by a centralized node or in
distributed manner.
High Each cloud data center is Each fog node can be equivalent to a single
Size very large in size consisting of at server machine. It’s designed to meet the
least thousands of servers. user demands.
Require sophisticated Depends on the environment. Majority of
Deployment
deployment planning them don’t require intense planning.
Operated in environment where it primarily
Operated in a fully controlled
depends on user demands. They are not
Operation environment with technical
operated directly by a person. It can be
expert teams by large companies.
operated by any size of company.
It can support predominantly
Can support cyber-domain, cyber-physical
cyber-domain applications. The
Applications applications. It suffers very less latency and
applications mainly suffer high
hence useful for time critical applications.
latency.
Require clients to have network Can operate autonomously to provide
connectivity until the user wants uninterrupted network services even with no
to access its services. Bandwidth or intermittent network connectivity.
Network requirements
requirement grows with the Bandwidth requirement depends on the total
increase in total amount of data amount of data need to be sent to cloud after
generated by all the clients. filtered by fog.
Table 3.1 Comparison of Fog and cloud

CONCLUSION between the demands of IoT infrastructure and the

computational capabilities cloud could provide. By
Fog computing has benefits in many domains, and creating a harmony between these two
provide the solutions for security problems in current complementary techniques of remote computing,
paradigm. Future research can expand the paradigm various targets can be achieved, which are once just a
of fog computing in smart grids. This provides us a dream for us.
vision to for fog to be a platform of unification made
enough for a brand new breed of rising services and REFERENCES
modify development of latest devices and
applications. However this study suggests that with [1] L. M. Vaquero and L. Rodero-Merino. Finding your way in
the fog: Towards a comprehensive definition of fog
the emergence of fog, cloud computing can not
computing. ACM SIGCOMM Computer
become an obsolete concept. As it was seen that, even [2] Yu Cao, Songqing Chen, Peng Hou and D. Brown, "FAST: A
if fog computing continue to develop in future, it has fog computing assisted distributed analytics system to
to go hand-in-hand with the cloud computing. Fog monitor fall for stroke mitigation," 2015 IEEE International
Conference on Networking, Architecture and Storage (NAS),
computing is just a bridge which connects the gap Boston,MA, 2015, pp. 2-11.

Fog Computing in IoT

4
 International Journal of Soft Computing and Artificial Intelligence, ISSN: 2321-404X, Volume-6, Issue-1, May-2018
http://iraj.in
[3] J. K. Zao et al., "Augmented Brain Computer Interaction [7] I. Stojmenovic and S. Wen, "The Fog computing paradigm:
Based on Fog Computing and Linked Data," 2014 Scenarios and security issues," 2014 Federated Conference
International Conference on Intelligent Environments, on Computer Science and Information Systems, Warsaw,
Shanghai, 2014, pp. 374-377. doi: 10.1109/IE.2014.54. 2014, pp.1-8. doi: 10.15439/2014F503.
[4] Zhu, Jiang, et al. "Improving web sites performance using [8] OpenFog Reference Acrhitecture, OpenFog Consortium,
edge servers in fog computing architecture." Service Oriented 2017 [Online]. Available:
System Engineering (SOSE), 2013 IEEE 7th International https://www.openfogconsortium.org/wp-
Symposium on. IEEE, 2013. content/uploads/OpenFog_Reference_
[5] “Top 5 reasons why the IoT needs Fog computing”, OpenFog Architecture_2_09_17-FINAL.pdf [Accessed: 4 November
Consortium, 2017 [Online]. Available: 2017].
https://www.openfogconsortium.org/5-reasons-iot-needs- [9] M. Chiang and T. Zhang, "Fog and IoT: An Overview of
fogcomputing/ [Accessed: 5 November, 2017]. Research Opportunities," in IEEE Internet of Things Journal,
[6] Amir Vahid Dastjerdi, Harshit Gupta, Rodrigo N. Calheiros, vol. 3, no. 6, pp. 854-864, Dec. 2016. doi:
Soumya K. Ghosh, Rajkumar Buyya, Fog Computing: 10.1109/JIOT.2016.2584538
Principles. Architectures and Applications, Cornell [10] Chih-Tien Fan, Zong-You Wu, Che-Pin Chang, Shyan Ming
University Library [Online]. Available: Yuan, "Web Resource Cacheable Edge Device in Fog
https://arxiv.org/abs/1601.02752 [Accessed: 5 November Computing", Parallel and Distributed Computing (ISPDC)
2017]. 2016 15th International Symposium on, pp. 432-439, 2016.



Fog Computing in IoT