Sai Vatsavai-Survey-Paper

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

A Survey Paper on Edge Computing - Fog

Computing,Mobile Edge Computing,Cloudlets

Sai Hima Varsha Vatsavai


Department of Computer Science
VCU, Virginia, USA
Vatsavaishv@vcu.edu

Abstract—Cloud computing has been a promising complexity, there is a massive growth in intelligent and mobile
solution for storage, computing and analysis of complex data devices coupled with technologies like Internet of Things,
for quite some time. The rapid boom of Internet of Things Augmented Reality etc. This demanded in gaining real-time
(IOT) devices and technologies, pose problems like high responses quicker and support context-awareness and mobility.
latency and low efficiency for time sensitive and location Hence, by looking into the available information it becomes
specific cloud applications/services, which need careful obvious that the main backbone of Cloud Computing is a data
consideration during design. This leads us to look at center which handles storage and processing of data. The
alternative design paradigms like FOG computing, multi- current paradigm of cloud computing architecture is to have
datacenters – capable of handling, storing and processing of
access edge computing and cloudlets, some of which seem to
massive amounts of data. Usually these data centers are
be promising down the line. In this paper, we shall discuss
connected to each other over WAN or Optical networks to
about the major alternative in the field of edge computing form data center networks (DCNs) [6]. These DCNs appear as
like Fog Computing, Cloudlets etc. We would also study the a single resource to the end user. They are usually suitable for
architecture, principles and implementation of edge low latency communication. However, with the advancement
computing technologies. In the later part of the paper, we in technology, data and Internet of Things (IOT) systems, they
will investigate the security and privacy issues related to the have demonstrated a new set of requirements to the existing
edge-computing platform and its effects on system cloud-based solutions. As it becomes increasingly difficult to
performance. Towards the end, we shall discuss about the manage the size and speed with which data is generated in
challenges and the research work being currently in today’s digital age, cloud computing can help us in building the
progress on these technologies. Finally, we will conclude by next generation of digital solutions.
discussing what can be achieved with current edge
Especially with the substantial increase in IOT in the last
computing platforms and how emerging technologies would few years there has been a lot of demand for decreasing the
handle the ever-increasing complexity of data and time latency and to improve efficiency. Also, the amount of
computing devices. This paper provides a complete insight data generated by sensors, actuators and other devices in the
that would be necessary in the field of edge computing. Internet of Things (IoT) has increased noticeably in the last few
years [12]. Generally, IoT data are processed in the cloud,
Keywords— Edge computing, Fog Computing, Cloudlets, mostly through computing resources located in distant data
DCNs, Internet of Things centers. Consequently, network bandwidth and communication
latency has become one of the serious problems. So, it is made
I. INTRODUCTION clear that even though the DCNs are suitable for low latency
The Computational and Storage world has become very communication, the latency of communication between the
large and complex. Cloud is an emerging technology in the end/edge devices and the DCNs must be monitored. Also, there
field of information technology. It is a scenario where in which is a need for mobility support due to the agile nature of the end
computing resources are offered as a service. It usually relies devices in several applications due to the advancement in
on sharing a physical or virtual resource rather than relying on present day technologies. There are also several works which
personal hardware or software. Many businesses are moving show the increase in connected devices in near future.
towards cloud to reduce their storage costs and to improve the There is a research which is conducted to estimate the
performance. When it comes to the matter of storage and number of connected devices in future. It states that the number
computation of large scales of data, cloud computing is the of connected devices would reach to around 30-50 billion by
better solution. It has significant potential to improve the 2020 [26]. Hence, this rate of increase would cause massive
deployment and administration of services through efficient scale of network traffic towards DCNs and connected devices
resource sharing and rapid scalability. which is a bottleneck degrading the latency and Quality of
The demand for cloud computing is increasing extensively Service and Quality of Experience. The other factor is, with the
with the increase in data and devices. Along with this data increase in requests to DCN leads to its operation at high duty
cycle. This effects the environment also due to the emission of wearables and gadgets, whom we refer as edge devices, and
harmful greenhouse gases. CDNs. This results in the increase in real time response.
Hence, there is a need to look ‘beyond the clouds’ towards the
The Edge Computing technology seem to overcome all the edge of the network. Hence, the paper [26] identified few
above-mentioned challenges faced in Cloud Computing necessities that made us look for edge computing which are
environment. Edge computing appears to tackle the challenges listed as follows:
and difficulties caused due to centralized computation, storage,
and networking in small datacenters, and due to the long- • To Decentralize Cloud
distance communication between the edge devices and the Centralized computing is not always the best idea for
remote datacenters. It appears to be the promising solution applications which are distributed. It’s always better when
which provides cloud resources closer to the resource-poor computing is done closer to the DCNs. So, it would always be
edge devices.
a better option to move the computing to edge servers.
According to Edge Computing, the computing resources Decentralizing cloud is beneficial to improve the performance
are made available at the end of the network, close to (or even of system.
co-located with) end-devices. By placing the computing • To Overcome the resource Limitation of End devices.
resources in such proximity to the devices reduces Front-end devices like smart phones, desktops have
communication latency. Furthermore, the data can be relatively less hardware to perform complex computations. So,
processed and analyzed just one hop away from end-devices, the data usually is sent to cloud to perform any sort of
which reduces the bandwidth demands on network links to
computation. It is not always a good practice to send all the
distant data centers. So, we can say that with the presence of
data to cloud not knowing which is required for any
these “Edge Devices”, the computational load at the data
centers is reduced drastically by handling or processing some computation. So, if the data is utilized in the edge servers
requests locally which do not require intervention from the instead of Cloud, this idea would always be better.
cloud. This reduces latency in resolving requests and handles • Dealing with Energy Consumption.
requests effectively. Finally, with all the knowledge we have With the increasing demand of cloud computing many
we can say that edge computing platforms support mobility of applications are moving to cloud platforms. This demand
devices and distributed applications. IoT deployments are also results in large energy consumption at the centralized cloud.
benefited from edge computing. So, with the diversified data centers this energy the
applications also get distributed to reduce energy
consumption.
• Dealing with Network Traffic.
Volume of data generated in today’s world is increasing
drastically. In-order to handle the incoming request to the
cloud there must be a concrete solution. Edge computing
handles network traffic effectively.
• Smart Computation Techniques.
Data generated at the user end needs to be sent to the cloud
server to perform any computation. There is a chance to
improve the computational performance if the data is sent to
nearby edge servers rather than sending to a distant cloud.
III. RELATED SURVEY
There are several other existing related studies in Edge
Computing Technology area that have attempted to provide a
survey of the papers in the field of Fog Computing, MEC and
Cloudlets. As discusses earlier, it is implemented in three
Figure 1: Architecture of Edge Computing ways which we are going to discuss in the coming sections.
The Edge Layer between the end devices and the Cloud are Fog computing presents M2M gateways and wireless
implemented in different ways based upon the communication routers as a computing layer between the actual cloud which
protocols, networks used, and services offered by it. It is are called Fog Computing Nodes (FCNs). They store and
implemented in three ways namely Mobile Edge Computing compute data from end devices locally before transmitting it to
(MEC), Fog Computing (FC), and Cloudlet Computing (CC). the actual distant cloud. Coming to MEC, it proposes the
The objective of the paper is to provide the survey, challenges deployment of intermediate nodes with storage and computing
and opportunities in edge computing. capabilities in base stations of cellular networks. With this
proposal it offers cloud computing capabilities inside the
II. MOTIVATION Radio Area Network [6]. On the other hand, Cloudlets are like
Using Cloud as the centralized server increases the number datacenters but on a lower scale near end users. This idea
of requests between users such as smartphones, tablets, allows end devices to perform their computation at the
cloudlet itself which acts as a resource similar to that of data • “Fog computing is a scenario where a huge number
center. of heterogeneous (wireless and sometimes
The authors in [27] present a comprehensive review of autonomous) ubiquitous and decentralized devices
current literature in Cloud computing with a focus on communicate and potentially cooperate among them
architectures, opportunities and challenges of cloud system. and with the network to perform storage and
The authors in [6] gives us the better understanding by processing tasks without the intervention of third
comparing all the three existing edge computing parties. These tasks can be for supporting basic
implementations. They concentrated in discussing about the network functions or new services and applications
parameters which are to be considered for better that run in a sandboxed environment. Users leasing
implementation of the edge computing. part of their devices to host these services get
The authors in [12] have discussed about the how incentives for doing so.” [4]
beneficial it is to bring the concept of computing closer to the
end users. The paper talked about the key IoT application
scenarios that benefit from edge computing and demonstrated • “Fog computing is a distributed computing platform
several experimental results which prove that edge computing where most of the processing will be done by
is necessary to achieve satisfactory improvement in the virtualized and non-virtualized end or edge devices.
performance. It is also associated with the cloud for non-latency-
The paper [18] discussed about the implementation of fog aware processing and long-term storage of useful
computing and about present security and privacy issues data by residing in between users and the cloud.” [2]
entitled with the architecture. Though fog computing resolves
many issues related to traditional cloud computing there are
many threats to the present architecture which are to be
addressed. The authors of the paper [19] have identified
twelve critical security issues and suggested possible solutions
which are discussed in the later part of the paper.
The paper [8] clearly explained about the architecture and
the current trends in the field of mobile edge computing. There
are number of surveys in the field of MEC which we have use
in this article. The paper [22] spoke about the cloudlet
technologies. It introduced Cloudlet mesh to secure the mobile
clouds from intrusions and attacks.
IV. IMPLEMENTATIONS
In this section, we will be discussing about the
implementation of the three edge computing technologies. We
critically examine the different implementations and compare
the features of these implementation. We would discuss about
the function and location of their nodes which serve as the
intermediate layer between the end devices and the actual
cloud, their target end devices and about their services offered
in coming subsections.
A. Fog Computing
There are various definitions for fog computing. We
already know that fog computing is a distributed computing
implementation where processing or computations are done at
the edge of the network with the integration of the actual cloud
Figure 2: Edge Technologies
through fog nodes. But, according to the authors of the paper
[2] fog computing is defined in various ways. Some of the
examples mentioned in this paper are as follows: Fog Computing extends Cloud Computing to the edge of
network, wireless networks for IOT applications. Fog
• “Fog computing is a highly virtualized platform that computing offers strong support for Internet of Things by
provides compute, storage, and networking services decentralizing the cloud environment and by distributing the
between IoT devices and traditional cloud computing computing and application services in most efficient way from
data centers, typically, but not exclusively located at the data source to the cloud. As lower amount of data is
the edge of network.” [2] transported to the cloud for processing, analysis and storage,
this implementation is highly efficient. The implementation of
fog computing is mainly based on the Fog Computing Nodes
(FCNs) which are placed between the end devices and cloud.
The FCNs are based on different kinds of elements like
routers, switches, access points, IOT gateways, set-top boxes
etc. The characteristics are well explained in this paper [5].
Characteristics of fog computing:
• Fog computing nodes are located at the network edge
and far away from actual cloud.
• They enable low latency.
• They are accessed over wireless network.
• Can support mobility of device and IoT.
• Support for on-line analytics with the cloud.
• Real-time interactions and Scalable.
Figure 3: Fog Computing architecture
Architecture of fog computing:
With the benefits of fog computing, we assume to have
a standard architecture. But, there is no standard architecture Components of Fog Computing architecture:
available till date. However, many research works have been The paper [2] well explained the components of fog
done in the field of fog computing. computing architecture consisting of several layers which are
The authors in this paper [3] well explained about the divided into groups depending on a functionality.
architecture of fog computing. According to this paper, According to this paper[5], the physical and virtualization
architecture is nothing but arrangement of physical and logical layer involves different types of nodes such as physical nodes,
network elements, hardware, and software to implement a virtual nodes and virtual sensor networks. While at the
useful network. The key architectural decisions mainly monitoring layer, resource utilization, the availability of
involve how the fog nodes are placed and the protocols, sensors and fog nodes and network elements are monitored.
topology, bandwidth capacities of links between the fog The pre-processing layer performs data management tasks.
nodes, devices and the actual cloud. It also involves the The pre-processed data are then stored temporarily in the
hardware and software of fog nodes. In order to optimize the temporary storage layer. In the security layer, the
efficiency and performance of fog nodes, one must understand encryption/decryption of data comes into play. Finally, in the
the critical requirements that will be using fog and specific transport layer, the pre-processed data are uploaded to the
software applications. Because, based on this study only the cloud to allow the cloud to extract and create more useful
requirements are mapped to a particular fog node. There will services.
also be certain requirements which are difficult to study or
implement and move to fog-based architectures as they are
heavily reliable on cloud. So, understanding these
requirements allows us to effectively implement and meet the
requirements of applications.
High-Level Architecture of Fog Computing:
This architecture comprises of three layers namely cloud
plane, fog plane, IoT plane which are represented in the below
figure as cloud, fog and infrastructure .
The most important layer is the Fog layer.It consists of
various computing devices mainy Fog Computing nodes
(FCNs). The problem to handle comple data is well solved by
handling the data at edge level. This FCNs are capable of
performing small and medium scale big-data processing.
The bottommost layer is Iot plane or infrastructure which
consists of all connected devices. For the time-sensitive
applications the processing has to be done in these FCNs, in
the fog plane and the other processing which is not time-
sensitive is done in the distant cloud. This fog layer usually
manages this requirement. It decides on which data to send to
coud and what should not be sent. With this implementation
the user is able to use the services of both fog and cloud based
upon the request. But, the complex processing and storage Figure 4: Components of Fog Computing
requirements are handled by distant cloud.
B. Mobile Edge Computing Characteristics of MEC:
Recent trend in smartphones or tablet computers has a • On-premises: MEC platforms can run isolated from
diverse effect on mobile and wireless networks [7]. Mobile rest of the network while accessing the local
networks have low storage capacity, high energy consumption, resources.
low bandwidth and high latency. Growth in IoT led to the • Proximity: MEC has the advantage to analyze big
increase in mobile networks in many ways. The main purpose data over normal computing technologies as it is
of MEC is to deal with the challenges from MCC. This deployed close to end users.
provides the end-user with faster complex computing, energy • Lower Latency: As MEC services are closer to user
efficiency, storage capacity, mobility when compared to the devices, we can achieve low latency due to less
computation in distant cloud. traffic.
Mobile edge computing is well defined in this paper [6], as • Location Awareness: MEC usually receives
an implementation of Edge computing to bring complex information from edge devices within the LAN to
computations and storage to the edge of the network to reduce know the location of the devices. This is one of the
latency and to improve context awareness. According to the important features of mobile networking
authors of [7], MEC is a convincible architecture where in
which cloud computing services are usually moved to the edge Architecture of MEC:
of networks leveraging mobile base stations. This To know the architecture of MEC this paper [8], tells us
implementation can be use in mobile, wireless and wireline the importance to learn about the cellular network
scenarios, which uses software and hardware platforms communication as MEC is usually found within RAN.
located at the edge closer to the end users. Role of RAN in Cellular Networks:
According to the paper [8], MEC is defined as: RAN is part of the cellular network communication system
• “Mobile edge computing provides an IT service infrastructure which acts as a connection between mobile
environment and cloud computing capabilities at the phones or any wireless machine with mobile core network
edge of the mobile network, within the radio access (actual network for data processing).
network (RAN) and in close proximity to mobile Similar to Fog Computing architecture, MEC architectures
subscribers.” is also classified into three layers. According to this paper
According to this theory, MEC offers cloud computing [10], MEC is a layer between the cloud and mobile devices. It
capabilities within RAN. This network reduces the traffic is an infrastructure where the data storage and processing
between main cloud and the end user, by connecting the end occur outside the end-user, in the edge devices.
user directly to the edge network. With the help of this we can According to the paper [6], MEC can be defined as the
improve our computational speed, avoiding bottlenecks and implementation of edge computing which brings the complex
system failure. computations and storage capacities to the edge of the network
The characteristics of MEC are well explained in this within RAN to reduce latency and to improve performance
paper [9]. and context awareness.
C. Cloudlets
In the paper [11], Proposed by Carnegie Mellon
University, cloudlet computing is another direction in mobile
computing that shares many traits with Mobile Cloud
Computing (MCC) and Mobile Edge Computing (MEC). It
deals with the disadvantages of MCC proposing a better
solution to improve the performance of the applications.
Various definitions of Cloudlets according to this paper
are:
• “A cloudlet is a trusted resource-rich computer or a
cluster of computers with strong connection to the
Internet that is utilized by nearby mobile devices”
[11]
• “A Cloudlet can be treated as “data center in a box”
running a virtual machine capable of provisioning
resources to end-devices and users in real time over
a WLAN network” [6]
Architecture of Cloudlets:
Cloudlet is the middle tier of a 3-tier architecture
according to [6] [11]. cloudlets are nothing but a small cloud
close to mobile devices, which provide one-hop access to the
Figure 5: MEC architecture
services with high bandwidth and low latency. The Hypervisor can run and create virtual machines that
architecture for Cloudlets is based on three layers, the run tasks. We can say that VM is an abstraction of a
component layer, the node layer and the cloudlet layer [6]. physical hardware and requires additional libraries
Cloudlet technology has been introduced to deploy mobile for hosting applications and services. Coming to the
cloud services, but it was inadequate because of the limited containers, they partition the resources of the
WIFI coverage. So as the computationally intensive tasks are physical machines by creating a multiple isolated
offloaded from the mobile devices, cloudlets can efficiently user-space not like VMs. We came up with Docker
perform computation reducing latency. which is the most significant container solution
facilitating edge computing technology.
• Network Function Virtualization (NFV)
It is a technology that allows operators to
function by separating network services and
functions from hardware. This allows operators to
function on software thereby increasing the speed and
performance.
The NFV architecture defines the following
domains:
1)VFNs, which are nothing but software
implementation versions of network functions.
2)NFV infrastructure (NFVI) which comprises of
hardware and software components creating a
network environment where VNFs are deployed.
3)NFV Management and Orchestration (NFV
MANO) which organizes and manages the physical
Figure 6: Cloudlets architecture and virtual resources.
• Software Defined Networks (SDN)
Various other papers also involved in the study of SDN is a technology which allows the multi-
cloudlets. They mentioned cloudlets as a small-box data center tenancy support. It helps in reducing the congestion
which are deployed at one hop away from mobile devices, in the core network by focusing on network
such as public places, office building etc. [12]. Several units of programmability, efficient resource sharing and real-
multicore computers form a cloudlet, connecting the end user time network control.
and the distant cloud. Cloudlet is mainly brought as a solution • Network Slicing
to overcome the latency, energy consumption due to the Network slicing is the technology which allows to
distant network by utilizing the connectivity to closer edge support businesses with different service
data centers. It basically utilizes technologies like WIFI requirements in the better way. It mainly slices the
located one hop or multiple hops away at the edge of the network into many instances making it compatible
internet. Therefore, this makes us aware that cloudlets are for a specific requirement or application or service. In
dependent on robust and uninterrupted internet connection. other words, we can say that network slicing gives us
This internet dependency will rise to several other issues like a multi-tenant environment which supports resource
security and privacy threats. sharing with the help of network slice broker
improving the performance and efficiency of the
V. ENABLING TECHNOLOGIES system. This service requires coordination between
The edge computing platforms which are described in the NFV and SDN technologies.
previous sections, can be achieved by a few enabling
VI. EDGE COMPUTING FOR IOT APPLICATIONS
technologies [12]. A related and detailed survey on enabling
technologies of edge computing is provided in [23]. There are IoT is the main reason for digitalization of the today’s
certain things which are important and can be used at edge world. It comprises of resource-constrained devices like
nodes. In this section, we will be discussing about some of the sensors, smartphones, wearable device etc. [12]. Several
important enabling technologies which are discussed in paper features enable IoT devices to be able to be deployed in edge
[23]. computing platforms rather than the distant cloud. The current
• Cloud Computing, VMs and Containers state of the art in the IoT largely involves device driven
Cloud Computing offers various deployment communications and does not require explicit human
model like private cloud, public cloud, hybrid cloud, intervention.
community cloud and various service models like As IoT has become a part of our lives, there is a rapid
IaaS (Infrastructure as a service), PaaS (Platform as a growth in data produced and services requested by them [11].
service), SaaS (Software as a service). We can say IoT generates massive amount of data. It becomes very
that cloud computing is currently in its mature phase. complex to transfer data to the cloud for computation. So, in
order to handle such data complexity, it is always better to addresses all these issues. Mobile operators host smaller
move the computation to edge devices rather than the distant servers at the premises of their base stations which brings the
cloud. storage and processing closer to the users.
A. Fog Computing for IoT
Fog computing bridges the gap between the cloud and
IoT devices by enabling computing, storage, networking, and
data management on the network nodes within the close
vicinity of IoT devices [11]. Hence with this technology the
data computation is done between the transformation of data
from end-users to the Cloud.
According to the paper [5], To increase the efficiency of
IoT applications, most of the data generated by these IoT
devices must be processed and analyzed in real-time. Fog
computing will bring cloud networking, computing and
storage capabilities down to the edge of the network, which
will address the real-time issue of IoT devices and provide
secure and efficient IoT applications [13].
Fog computing is considered to be the best practice for
applications with low latency requirements such as video
streaming, gaming, augmented reality, etc., according to this
paper. The integration of fog computing with the IoT will
bring many benefits to various IoT applications.
Other related papers have focused on the security and
privacy issues also, which we will be discussed in further Figure 8: Applications of MEC in IoT
sections. Potential applications for MEC-enabled IoT include smart
mobility, connected vehicles, emergency response, smart
cities, content distribution and location-based services. After
looking into the idea of MEC we can say that, it is the best
alternative solution for a movable edge device.
C. Clodlets for IoT
We already discussed about how the usage of number of
mobile devices is increasing. These devices are connected to
the internet and sending huge amount of data to the cloud
which increases bandwidth and results in high latency. [15].
To address this issue Cloudlets have come into picture.
Cloudlets are nothing much a micro-cloud closer to the end
users.

Figure 7: Applications of Fog Computing in IoT


B. Mobile Edge Computing for IoT
The paper [14], explains how IoT is welcoming MEC
capabilities, improving the performance and capabilities. So,
we are extending the already existing works in various other
implementations to overcome the concerns like scalability,
privacy and security threats. Hence a new architecture of IoT
is proposed to overcome the cited challenges.
Cloud platforms are mostly not suitable for applications
Figure 9: Applications of Cloudlets in IoT
which involved low latency requirements, low responsive
time, high performance and high quality of service. MEC
Cloudlets usually provides its services to the users closer • Spam: Spam results in unwanted usage of data,
to LANs. Cloudlets are nothing, but virtual machines which resources and energy.
are in one-hop proximity to the user devices. They provide • Jamming: attacker might generate large amounts of
mobility, scalability and elasticity to many of the user fake data to jam a transmission channel.
applications. • Denial of service: attackers might flood the fog
server with fake requests making it unavailable to the
The paper [15], concluded saying that this implementation actual users.
of the cloudlets is one of the best practices in edge computing. • Collusion: Two or more attackers collude to
In this ideology the user requests the service from the distant strengthen and to degrade the performance of cloud.
cloud through cloudlets. With this kind of implementation, we • Insecure Interfaces: Relying on insecure interfaces
can say that the users will obtain the requested services from and APIs lead to security issues. If we understand the
the cloud in less time compared to the situation where there is security model of cloud provider and the dependency
no cloudlet and where the computation has to be done directly chain associated with API, we can avoid such
in the distant cloud. The main limitation of user devices is the incidents.
energy consumption. This issue can be handled by cloudlet as • Cookie poisoning: This involves unauthorized access
cloudlet is connected to the cloud and it allows faster response to the application by changing the contents of the
to the service requests to or from the cloud and thus limiting cookie. We can erase data in the cookie or encrypt
the energy consumption. The cloudlet technology is thus used the data involved in the cookie to avoid the issue.
in various fields like education, crowd management, banking Which user wants his data to be available to all the other
and business and E-health systems. users. Cloud offers data storage, that means users data is
VII. SECURITY AND PRIVACY THREATS available in cloud. Cloud providers must make sure that the
user’s data is private and secure. Privacy is also one of the
Now a days many businesses are moving to cloud most important criteria to consider. We have user’s privacy
computing for efficient storage management and to improve divided into four aspects namely Identity privacy, Data
efficiency. In order to handle the traffic to the distant cloud, privacy, Usage privacy, Location privacy.
the idea of edge computing has come into picture. In this paper [11], the author not only tried to address
Edge computing moves the complex computations to the various threats effecting fog computing but also suggested
edge of the user devices instead of the main cloud server. With various solutions and protocols to improve current available
this demand for cloud, security and privacy threats are also protocols.
increasing. Security and privacy threats are the main According to this paper [19], Cloud security alliance have
limitations to edge computing implementation. identified twelve critical security issues and suggested
Though many researches and cloud providers are working possible solutions. Let us look into some of the issues.
on it to avoid such attacks, but these incidents keep on
happening. Many incidents on AWS, Dropbox, Drive etc.
Possible
prove that cloud computing is not immune to security and Attack Category Possible Threats
Solutions
privacy threats. Security is one of the most difficult tasks to Multi factor
implement in edge computing technology. Virtualization Hypervisor attacks, privilege Authentication,
issue Escalation attacks Control mode
access
A. Security and Privacy issues in Fog Computing Antivirus
protection,
The paper [18], also clearly discussed about the security Web Security Foreign session attack,
Intrusion
issues Hijacking
issues in fog computing. It mentions that cloud is vulnerable to Prevention
many security threats because of its computing framework and System.
Encrypted
centralized data storage. It draws the idea that since fog Internal or
Insecure APIs, Inefficient Communication
inherits many features from cloud, it inherits risks also. Communication
policies or multifactor
Issues
Let us discuss about the few threats in fog computing Authentication
implementation which are discussed in this paper: Anti-malware
Malware Virus, malware functions, functions,
• Abuse of cloud computing: Cloud computing is a functions trojan Rigorous data
week entity. The misuse of cloud technology can be backup
avoided by strictly validating and registering the Wireless security Distortion issues, data loss, Communication
actual users of the cloud. issues data breach, sniffing attacks key
• Forgery: Attackers try to imitate their identity to Policy
access other information which results in Data Security
Data replication, Data enforcement
performance degradation due to fake data sets. altering, Data erasing security,
related issues
attacks architecture
• Tampering: To affect the performance attackers encryption.
might modify the data that is being transmitted.
TABLE I. THREATS AND SOLUTIONS
encryption can secure users information which is discussed in
paper [20]. It uses a super increasing sequence to structure
multidimensional data and encrypt by the homomorphic
cryptosystem technique.

Figure 10: Threats in Fog Computing

B. Security and Privacy issues in Mobile Edge Computing


In the era of mobile technology many applications are Figure 11: Threats in Mobile Edge Computing
moving towards MEC to improve their performance by
transferring the complex computations at the MEC nodes Another privacy issue is user’s location. GPS has is both
distancing the main cloud. MEC has come to the picture with boon and a bane. With the help of GPS attackers will be able
applications in video analytics, big data analytics etc. to track the location of the users. In order to address this issue
The paper [7], has provided with all the details of security MobiShare system is widely used.
concerns and necessary solutions to avoid such incidents. The The author of the paper [17], concluded that having
three main aspects which are to be considered while dealing encryption module at mobile device end and HSM (hardware
with security and privacy threats in MEC architecture are: security module) at the cloud end will be the better solutions
Confidentiality, Integrity and Availability. to information security and data privacy.
• Network security is very important element in C. Security and Privacy issues in Cloudlets
Mobile Computing. It becomes very difficult to
We all know that Cloudlet is a reliable, resource rich
mitigate the attack as the data has to be transmitted
cluster which operates only when connected to the internet.
from user and cloud and this transmission layer has to
This dependency makes it not immune to the security and
be secured.
privacy threats.
• Core network security is again the important aspect The author [21] proposes a data security protocol which
to considered. Edge devices usually exchange not only protects the data from unauthorized user but also
information and bypass the central cloud security prevents exposure of data to the cloud owner. It offloads all
mechanism. Hence, MEC is prone to attacks. the complex computations on the cloudlet and allows
• In MEC server security, there may be attacks which lightweight wireless devices to securely store and retrieve
hijack the whole data center. Design flaws, their data in the cloud.
Configuration errors, Insufficient security One of the interesting proposals in the field of edge
managements might be the reasons to such attacks. computing is cloudlet mesh architecture [22], which provides
• It is important to have proper end devices. Improper security enforcement to establish a better cloud environment
selection of end devices might result in the damage in not effected by security and privacy issues.
the performance of whole system. According to this proposal, the cloudlet mesh is a WIFI or
Hence, to address all these issues, security measures are mobile-connected over the internet which acts as a shield
important [7]. Authentication is one of the important measures against the spam, virus, worm attacks and stops unauthorized
to protect the MEC environment from such threats. To handle access to the data in cloud.
Network security attack, Intrusion detection systems can be
employed in MEC data center. TABLE II. THREATS AND SOLUTIONS
Virtualization and physical servers can be protected
through hypervisor hardening, network abstraction and Threats or attacks Cloudlet
isolation policies. Virus, worms, malware Protects by verifying files
Privacy is again an important factor to be considered. It is attacks and content used by the user
Identity theft and Need to authenticate cloudlet
mandatory to secure user’s data, location, usage etc. This is authentication user, public cloud also
quite challenging. Aggregation schemes such as homomorphic
Threats or attacks Cloudlet VIII. FUTURE RESEARCH WORK
Cloud offloading and file Data caching at cloudlet
transfer improvising performance
Fog Computing, Mobile Edge Computing and Cloudlet are
the main edge computing technologies. Each one of them have
Various protocols have been proposed in [22], to secure different characteristics and offers different services to the
the end devices of the user and the central cloud which are as user. Much research work is going on to improvise these
follows: technologies and to develop new features to meet the needs of
• Multi-party Authentication Protocol: This is mainly different customers. With this mature state in cloud computing
used for mobility management, multi way environment there are many aspects to discuss. The main
authentication. aspect to consider is, there are no strict policies or standards to
monitor these technologies. So, it becomes clear that if the
• Inter-Cloudlet Protocol: This protocol can detect
policies are standardized, there would be a better clarity on the
unwanted operations with the help of cloudlet
features and information of edge nodes [6], giving a better
installed with Intrusion detection system (IDS). This
environment to the end user.
system protects the cloudlet from fake users by
Recently, Mist Computing has been introduced. It
maintaining a database which consists of actual users.
captures the most extreme edge of the connected or user
This prevents the fake users from involving in
devices [11]. This system is clearly proposed keeping future
communication between the user, cloudlet and distant
self-aware and automatic systems in mind. It is also named as
cloud.
IoT computing or Things computing. Mist computing can be
• Trusted cloud transfer Protocol: According to this
used to reduce the load in traditional WiFi infrastructures by
protocol, for every request from the user to cloud or
distributing the load accordingly between the nearest edge. It
vice versa, the cloudlet uses TCTP to encrypt the data
can preserve the privacy of user’s data by processing it in local
which is being sent to the distant cloud. The process
environment.
in which the TCTP handshaking protocol takes place
In today’s era of smart phones and tablets, to increase
id:
the speed and performance of the application all the mobile
1)Whenever a mobile user request for a contact of
applications are moving to cloud. This trend predicts that with
the cloudlet, Certificate Authority (CA) sends a
more and more devices moving to cloud there is a necessity of
public key to the cloudlet.
revisiting the already existing security and privacy controls to
2)The data which is transmitted to the cloud is
cope up with the current challenges. Hence, we can say that
usually encrypted. The cloudlet scans every packet of
the future work lies in establishing the standardized privacy
data and detects virus or spam. If anything is found it
and security controls within the edge computing technologies
issues an alert otherwise the data is sent to the device.
discussed in the paper to survive the coming advancements in
3)The packet will be sent back if the cloudlet
security needs.
hasn’t heard anything from the device. It notifies the
It is obvious that Edge Computing technologies require
device accordingly after receiving any report from
more sophisticated setup when compared to the pure cloud
the distant cloud.
computing systems. It involves many patterns and resources
These measures will really help in preserving the
which requires scientific scrutiny for the better performance.
security in the cloudlet.
In order to bridge this gap, a new simulator tool
EdgeCloudSim has been proposed [25]. In this paper the
author clearly discussed about the existing scrutiny techniques
like CloudSim, Omnet++, Opnet, NS-3 etc. and talks about the
existing works. CloudSim is a simulation tool which allows
modelling of cloud computing infrastructure and application
services in the standard cloud computing setup. Drawbacks of
CloudSim are lack of wireless LAN, lack of mobile nodes,
lack of realistic VMs etc. EdgeCloudSim addresses all the
above issues. It supports simulating multiple edge nodes
system where edge servers are running in parallel with the
central cloud.
EdgeCloudSim is available as an open source project
which allows its users to incorporate their needs in the tool.
The paper has demonstrated the working of the tool on various
architectures and environments and the results truly show
positive effect on the performance.

Figure 12: Threats in Cloudlets


IX. CHALLENGES Efficient management and utilization of the edge
Having discussed a variety of edge computing nodes, data packets and other resources results in
technologies and several research papers, we can learn about reduced power consumption.
the challenges faced in these technologies. Though Fog • Cost:
computing, Edge computing, Cloudlet, decentralize cloud The cost of each new technology influences the
computing, these technologies often rely on centralized cloud existing cost. The cost of integrating all the necessary
platform. Hence, these technologies not only draw the good new technologies to the existing architecture must be
things from the cloud but also the challenges faced in the monitored. Optimizing the cost is a real time
existing architecture. Let us discuss some of the important challenge. The idea is to develop a application with
challenges in edge computing environment [1] [2] [11] [24]: secure architecture and least cost. Hence, the
• Security and Privacy: developed architecture has to avoid the modules that
Security is not only one of the important consume power unnecessarily to optimize the
challenges faced in the present day but also remains functionalities.
the most important challenges faced in the future • Deployment Issues:
edge applications. Future edge cloud uses multiple Excessive increase in the layers while
technologies which also multiplies the security and implementing edge computing technologies, might
privacy issues. As edge cloud is nothing but the cause deployment issues. It might result in latency
extension of the actual cloud, it draws most of the problems. Deployment decisions must be taken
issues from the earlier technologies. Since the edge considering the requirements like how many sensors
clouds are closer to the end users they appear to be are needed, task division on edge nodes etc.
more fragile due to the frequent access from various Application and resource optimizing play a very
users. As edge cloud is an open platform it hosts important role while deploying a particular
various third-party hardware and software. So, environment whether it is fog computing or mobile
software security is also important. Confidentiality, computing or cloudlet.
Integrity and availability are the main principles of • Infrastructure related issues:
information security. With the accelerating advancements in edge
Appropriate authentication, Authorization computing technology there are many open
measure are necessary to identify the trusted parties challenges due to its interoperable architecture. This
and to defend unnecessary access request from open architecture needs to support all kinds of time-
miscellaneous sources which are not safe. This also sensitive applications.
protects the privacy of user’s data without being • Failure management:
public to all the users. The probability of failure occurrence is high in
• Coordination between IOT devices and the Edge: the edge computing environment because of its
The main concept of edge computing paradigm is distributed and decentralized nature. Software failure,
to deploy edge servers in proximity to the end users, Hardware failure or failure due to the user’s activity
IoT devices, between the cloud and the user. How the might be some of the reasons of failure. Beside these
edge servers and the connected devices coordinate problems there are some other issues like
will have impact on the performance of the system. connectivity, power source etc. Wireless
To address this issue VMs have come into picture. communication is the main mode of communication
Instead of sending the data to distant cloud for in the edge technologies and it is obvious that
computation, VMs can dynamically be launched in wireless communication is not reliable and is immune
the edge location and can be used for local to system failures. As most of the devices connected
computing. This reduces the cost of offloading to the to the edge clouds are run on power, we need to
cloud. ensure that their battery is monitored and do not fail.
• Quality of Service: • Communication between different layers:
Quality of Service normally comprises of quality Let us consider an example of automatic car. In
of connectivity, data delivery delay, response time case of emergency failure in connectivity might cause
etc. With the involvement of various parameters in severe harm. Hence, in such situations even if
the edge computing environment QoS is always connectivity to the distant cloud fails, the
questionable. Adhering to the rules and regulations of environment needs to ensure the connectivity to the
the system we can easily achieve the service quality. nearest edge server. Hence, the cross-layer
• Power Consumption: communication between the cloud, edge cloud, user
Optimizing the energy consumption in the edge is very important and how the connectivity issues are
computing architecture is a serious challenge to be handled is still a research topic.
considered. Involving many number of nodes, • Application related issues:
complex data collecting, transmitting and computing Multiple Applications run on the shared
increases the power consumption drastically. environment in edge clouds. Hence, this environment
is not immune to security and privacy attacks which
might degrade the performance of the cloud. So, it is
B. Experiments with Light Weight Virtualization
important to address the application-level security
Technologies
issues like appropriate application isolation, shared
traffic and data access for multiple applications. The authors of the paper [11] conducted experiments to
• Application modelling: evaluate three container tools namely, Google Kubernetes,
Modelling applications in edge computing Docker Swarm, Apache Marathon and to study their
technologies is quite complex because it involves performance for IoT applications. The experiment is
collection and processing of complex data. Other conducted based on three requirements. Firstly,
concern is to deploy those applications with minimal adding/removing a new node must be a seamless process.
efforts. To solve this issue, a standard communication Second, Specific node communication has to be enabled.
protocol is necessary. This allows the proper Third, all the resources available at the node must be
communication with the Cloud, User, IoT devices accessible by the container on top of it. They proposed a new
and Edge nodes. container based on Docker for fog computing environment
• Auditability and Forensics: that can meet all the above requirements
As data is not maintained locally, it is difficult to
audit data held outside the organization in the distant C. Experiments with High Computation Applications
cloud which is available to many other users.
Interactive applications are said to be challenging for edge
• Resource Monitoring: computing and cloudlets. The authors in [11] proposed a new
Monitoring plays a very important role when serverless architecture for MEC environments to reach high
there are multiple users for a specific application. So, throughput and keeping low latency. The experiment is
developing resource monitoring techniques are very conducted on mobile AR application, in which captured
much essential. frames from the camera must be analyzed to detect the object.
X. EXPERIMENTAL EVALUATIONS Various experimental studies on multi player online games,
machine learning packages, prove that the deploying
Several studies in edge computing area have conducted computational loads at the edge of the network can improve
experiments to verify the concepts and ideas experimentally. the performance drastically.
In this section, let us look into some of the experiments
conducted in the previous research papers: XI. DISCUSSION
The paper attempted to discuss about most of the important
A. Evaluation of Mobile Gaming using prototype Edge information one should know about the edge computing
Computing Platform technologies. Fog computing, mobile computing and cloudlets
Mobile gaming is one of the currently available share the vision of edge computing technology but each one is
technologies. Evaluating the edge technologies in this scenario different from one another in terms of architecture,
is meaningful. The paper [12] conducts the evaluation of edge implementation, customers, data handling techniques etc. Lack
computing in a GamingAnywhere platform. They used of standardization is clearly pictured in all the
Neverball game, where the user needs quick responses from implementations which effects the features and requirements
the server to have a good quality of experience. It uses a of the user. Hence, more research into this area would result in
mobile phone (end device) that sends the relevant game input better standards and implementation of edge computing.
to the gaming server, which in turn renders the content and Offloading computation to regional data centers reduces
streams the video back to the end-device. They considered network latency ensuring better quality of experience for the
three different server deployment scenarios: a local end users. So, deploying computing resources at the edge
deployment at the network edge, a special purpose cloud helps improve the performance of the system. It would be
computing infrastructure, and a commercial public cloud interesting if future research work is done to compare the
provider. performance of different edge computing platforms in
The experiment mainly concentrates on network delay, different application specific scenarios.
overhead of different technologies and the impact of edge Confidentiality, Availability and Integrity has to be the
computing on the processing delay. The results show that the promising services offered by the cloud. This allows many
network delay and overhead is more in the normal cloud when users to protect their data from intruder attacks. Recently we
compared to the cloud service which hosts edge nodes. This have seen several incidents on Facebook, AWS which clearly
proves that hosting computing resources very close to the end- show that the cloud is not immune such security attacks. Data
users, is the better option to achieve a satisfactory quality of handling seems to be the nightmare to the service providers.
experience Security is still a research issue. Few related researches in the
However, the experiments involved only one end-user. field of security and privacy is always encouraged.
Hence, more investigation has to be done to understand the MEC has the potential to emerge as the best edge
performance when it is connected to multiple users. computing technology, provided it concentrates more on
security & privacy issues, battery, efficient data handling and
open issues. We have discussed about the technology of [6] Koustabh Dolui, Sowmya Knati Datta. Comparision of edge computing
implementations: fog computing, cloudlet and mobile edge computing.
MEC, applications, research challenges and some issues int In Global Internet of Things Summit, 2017
his paper. Future work can be concentrated on designing [7] Yifan Yu. Mobile edge computing towards 5G: vision, recent progress,
scalable techniques which supports high speed users. and open challenges. In China Communications, pages 89-99, 2016
The below table depicts the information of the articles in [8] Milan patel, Jerome Joubert, Julian Roldan Ramos, Nurit Spercher,
various conferences which are used in this survey paper. Sadayuki Abeta, Adrian Neal. Mobile-edge computing—Introductory
technical white paper. 2014
[9] Tom H. Luan, Longxiang Gao, Zhi Li, Yang Xiang, Guiyi Wei, Limin
TABLE III. INFORMATION Sun. Fog Computing: Focusing on mobile users at the edge. 2015
[10] Ashkan Yousefpour, Caleb Fung, Tam Nguyen, Krishna Kadiyala,
Fatemeh Jalali, Amirreza Niakanlahiji, Jian Kong, Jason P. Jue. All one
Category Papers
needs to know about Fog Computing and related Edge Computing
Foundation [1] [2] [6] [12] [26] Paradigms: A Complete Survey. 2018
[2] [3] [10] [20] [27] [11] Gopika Premsankar, Mario Di Francesco, Tarik Taleb. Edge Computing
Framework and Architecture for the Internet of Things: A case study. 2018
Design and Implementation [2] [10] [12] [22] [12] Amir Taherkordi, Frank Eliassen. Data-centric IoT services
provisioning in fog-cloud computing. 2017
[2] [4] [7] [10] [11] [12] [14]
Applications [13] Ola Salman, Imad Elhajj, Ayman Kayssi, Ali Chehab. Edge computing
[21] [22]
[10] [15] [16] [17] [18] [19] enabling internet of things. In IEEE 2nd world forum of Internet of
Security and Privacy Things, 2015
Challenges [6] [10] [12] [23] [25] [14] Neenu Negi Rana, R. Rajkumar. Information harvesting @Intelligent
systems using cloudlet and IoT in cloud systems. In International
Evaluation [11] [24] conference on smart technologies for smart nation, 2017
[15] G. Racherla, D.Saha. Security and privacy issues in wireless and mobile
computing. In IEEE conference on personal wireless communications,
2002
[16] Debasish Jana, Debasish Bandyopadhyay. Efficient management of
XII. CONCLUSION security and privacy issues in mobile cloud environment. In Annual
Edge Computing has still the potential to offer more IEEE India Conference, 2013
efficient distributed computing. Though it is the best solution [17] Abdullah Aljumah, Tariq Ahamed Ahanger. Fog Computing and
security issues :a review. In ICCCC, pages 237-239, 2018
for storage and computing, there are many unsolved issues.
[18] Saad Khan, Simon Parkinsosn. Fog Computing security: A review of
This paper is a survey on different edge computing current applications and security solutions. In Journal of Cloud
technologies and implementations. It also presents a brief Computing Advances, 2017
study on security, privacy issues and the challenges in field of [19] Mithun Mukherjee, Rakesh Matam, Lei Shu, Leandros Maglaras,
edge computing paradigm. IoT is accelerating and is beneficial Mohamed Amine Ferrag, Nikumani Choudhury, Vikas Kumar. Security
and privacy in fog computing: Challenges. 2017
to many businesses. Due to the benefits of IoT, it is expected
[20] Manisha Jindal, Mayank Dave.Data Security protocol for cloudlet based
to connect with billions of devices and users. Hence, this architecture. In International conference on Recent Advances and
paper provided a taxonomy of research topics in edge Innovations in Engineering, pages 1-5, 2014
computing for IoT. The information provided in this paper [21] Yue Shi, Staampatoor Abhilash, Kai Hwang. Cloudlet mesh for securing
might help us in developing efficient edge nodes which enable mobile clouds from intrusions and network. In IEEE conference on
to improvise the performance of the application. The focal mobile cloud computing, services and engineering, pages 109-118, 2015
purpose of this paper is to summarize the edge computing [22] Tarik Taleb, Konstantinos Samdanis, Badr Mada, Hannu Flinck, Sunny
Dutta, Daario Saabella. On multi-access edge computing: A survey of
technologies. the emerging 5G network edge architecture & orchestration. In IEEE
Communications Surveys & Tutorials, 2017
[23] Imen Bouzarkouna, M’hammed Sahnoun, Nouha Sghaier, David
REFERENCES Baudry, Christian Gout. Challenges Facing the industrial
implementation of fog computing. In IEEE conference on future Internet
of Things and Cloud, pages 341-348, 2018
[1] Jilani Pan, James McElhannon. Future Edge cloud and Edge computing
for internet of things applications. In IEEE Internet of Things Journal, [24] Cagatay Sonmez, Atay Ozgovde, Cem Ersoy. EdgeCloudSim: An
Volume:5, Issue:1, pages 439-449, 2018 environment for performance evaluation of Edge computing systems. In
Second International Conference on Fog and Mobile Edge Computing,
[2] Ramesh Kumar Naha, Saurabh Garg, Dimitrios Georgakopoulos, Prem
pages 39-44, 2017
Prakash Jayaraman, Longxiang Gao, Yong Xiang, Rajiv Ranjan. Fog
Computing: Survey of trends, architectures, requirements and research [25] Blesson Varghese, Nan Wang, Sakil Barbhuiya, Peter Kilpatrick,
directions. In IEEE Access, pages 47980-48009, 2018 Dimitrios S.Nikolopoulos.Challenges and opportunities in Edge
computing. In IEEE International conference on Smart Clous, 2016
[3] Charles C.Byers. Architectural Imperatives of Fog Computing: Use
cases, Requirements and Architectural techniques for fog-enabled IoT [26] Shyam Paatidar, Dheeraj Rane, Pritesh Jaain. A Survey paper on Cloud
networks. In IEEE Communications Magazine, pages 14-20, 2017 Computing. In Second International conference on Advanced computing
and communication technologies, pages 394-398, 2012
[4] Luis M.Vaquero, Luis Rodero-Merino. Finding your way in the fog:
Towards a comprehensive definition of fog computing. In AGM [27] Luciano Baresi, Danilo Filgueira Mendonça, and Martin Garriga.
SIGCOMM Computer Communication Review, pages 27-32, 2014 Empowering low-latency applications through a serverless edge
computing architecture. In European Conference on Service-Oriented
[5] Hany F.Atlam, Robert J.Walters, Gary B.Wills. Fog computing and IOT
and Cloud Computing, pages 196–210. Springer, 2017
:A review. 2018

You might also like