This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI

Edge Computing in 5G: A Review

NAJMUL HASSAN1 , KOK-LIM ALVIN YAU (SENIOR MEMBER) 1 , CELIMUGE WU (SENIOR
MEMBER) 2
1
Department of Computing and Information Systems, Sunway University, Selangor 47500, Malaysia.
2
Graduate School of Informatics and Engineering, The University of Electro-Communications 1-5-1 Chofugaoka, Chofu, Tokyo, 182-8585, Japan.
Corresponding author: Kok-Lim Alvin Yau (e-mail: koklimy@sunway.edu.my).

ABSTRACT 5G is the next generation cellular network that aspires to achieve substantial
improvement on quality of service, such as higher throughput and lower latency. Edge computing is
an emerging technology that enables the evolution to 5G by bringing cloud capabilities near to the
end users (or user equipment, UEs) in order to overcome the intrinsic problems of the traditional
cloud, such as high latency and the lack of security. In this paper, we establish a taxonomy of edge
computing in 5G, which gives an overview of existing state-of-the-art solutions of edge computing
in 5G on the basis of objectives, computational platforms, attributes, 5G functions, performance
measures, and roles. We also present other important aspects, including the key requirements for
its successful deployment in 5G and the applications of edge computing in 5G. Then, we explore,
highlight, and categorize recent advancements in edge computing for 5G. By doing so, we reveal
the salient features of different edge computing paradigms for 5G. Finally, open research issues are
outlined.

INDEX TERMS 5G, cloud computing, edge computing, fog computing.

I. INTRODUCTION massive amount of real-time data, b) process, compute,

and analyze the data, and c) make and distribute deci-
Edge computing is a computational paradigm that en- sions on mini clouds locally. Hence, edge servers in the
ables edge servers in mini clouds (or edge clouds) to mini clouds have the capabilities of a cloud but on a
extend cloud capabilities at the edge of the network different scale, and they are located locally instead of
to perform computationally-intensive tasks and store a remote data centers which may be far away from UEs
massive amount of data at close proximity to user equip- [9].
ment (UEs) [1]–[3] . Traditional cloud computing, which
is a centralized computing paradigm that provides con-
A. OUR CONTRIBUTIONS
tinuous access to highly capable data centers, has been
adopted to allow UEs to offload computation and storage This paper highlights recent advances of edge computing
to the data centers [4]. This is because UEs have lim- in 5G. Some analyses have been made on a particu-
ited processing, computational, and storage capabilities. lar computational platform of edge computing in 5G,
Nevertheless, edge computing is preferred to cater for particularly mobile edge computing (MEC) [10], [11]. In
the wireless communication requirements of next gener- addition [12], [13] focuses on edge orchestration and its
ation applications, such as augmented reality and virtual related issues in 5G environment and MEC architecture.
reality, which are interactive in nature. These highly in- In [14] a survey of edge computing, including its applica-
teractive applications are computationally-intensive and tions and key challenges, is presented from the perspec-
have high quality of service (QoS) requirements, including tive of vehicular networks. Our paper is first of its kind to
low latency and high throughput (e.g, ultra reliable low present: a) a taxonomy of edge computing in 5G covering
latency communication (URLLC), tactile internet) [5]– the objectives, computational platforms, attributes, the
[7]. Most importantly, these applications are expected to use of 5G functions, performance measures, and the role
generate a massive amount of data up to 30.6 exabytes of edge computing; b) a review of state-of-the-art edge
per month [8]. The limited capabilities of UEs warrants computing schemes in 5G; and c) open issues in this
the need for edge computing to: a) receive and store a research topic. This topic is timely due to the recent

VOLUME 4, 2016 i

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

advent of 5G, and the evolving roles of edge computing sector) to provide directional transmission (or beamform-
in the realization of 5G. ing) in order to reduce interference, allowing neighboring
nodes to communicate simultaneously.

B. ORGANIZATION OF THIS PAPER

B. MAIN CHARACTERISTICS OF 5G DATA
The rest of this paper is organized as follows: Section
II presents an overview of 5G and edge computing, Data can be categorized into three main categories ac-
respectively, and answers a host of questions on the use of cording to its time characteristics as follows:
edge computing in 5G, including the time characteristics
• Hard real-time data has a strict predefined latency.
of data, the key requirements, and the applications of
Applications, such as video steaming, gaming, and
edge computing in 5G. Section III presents a taxonomy
healthcare services, generate this kind of data.
of edge computing in 5G. Section IV presents the state-
of-the-art schemes for edge computing in 5G. Section V • Soft real-time data has a predefined latency, yet it
presents open research issues. Section VI concludes this can tolerate some pre-defined and bounded latency.
paper. Applications, such as intelligent traffic signal control
system, generate this kind of data.
II. BACKGROUND • Non-real-time data is not time-sensitive and can
tolerate latency.
This section presents an overview of 5G, the time char-
acteristics of data, edge computing, as well as the key Edge computing is envisioned to handle applications
requirements and the applications of edge computing in and services with hard real-time requirement using edge
5G. servers due to their close proximity to UEs leading to
significant reduction in latency. For applications and
services with soft-real time requirement, or bounded end-
A. REQUIREMENTS OF 5G SYSTEMS
to-end delay, tasks are handled by edge servers if the
5G is foreseen as the next generation wireless cellular net- response delay between UEs and the cloud is higher than
work to cater for the needs of next generation networks. the requirement; otherwise, the tasks can be offloaded to
5G possesses three main characteristics unseen in pre- the cloud. For applications and services with non-real-
vious generation networks. Firstly, a massive amount of time requirement, tasks can be offloaded to the cloud
data is generated. According to the International Telecom- for load balancing.
munication Union (ITU), there are more than 7.5 billion
mobile devices around the world in 2017 [15], and the
C. SIGNIFICANCE OF EDGE COMPUTING
number of mobile devices is expected to increase to 25
billion by 2020 [16], contributing to ultra-dense networks. 5G is foreseen to support highly interactive applications
Consequently, there is an explosive growth in the amount with low latency and high throughput requirements [21].
of data from 16.5 exabytes in 2014 to an estimate of 500 Edge computing adopts a decentralized model that brings
exabytes in 2020 [17], contributing to a growth rate of 30 cloud computing capabilities closer to UEs in order to re-
times. Secondly, stringent QoS requirements are imposed duce latency. Fig. 1 shows the cloud computing and edge
to support highly interactive applications, requiring ultra- computing models. Edge computing can either operate as
low latency and high throughput. Thirdly, heterogeneous a single computing platform, or a collaborative platform
environment must be supported to allow inter-operability together with other components, including the cloud [22].
of a diverse range of UEs (e.g., smart phones and tablets), Edge computing is necessary as the traditional cloud
QoS requirements (e.g., different levels of latency and computing model is not suitable for highly interactive
throughput for multimedia applications), network types applications that are computationally-intensive and have
(e.g., IEEE 802.11 and Internet of things), and so on. high QoS requirements, including low latency and high
throughput. This is because cloud may be far away from
5G is comprised of three main new technologies to pro-
UEs, which also increases energy consumption. In other
vide higher network capacity in order to support a higher
words, cloud servers are typically located at the core
number of UEs [18]. Firstly, mmWave communication,
network, and edge servers of the mini clouds are located
which uses high frequency bands (i.e., 30 GHz to 300
at the edge of the network [23].
GHz [19]), provides high bandwidth (i.e., at least 11 Gbps
[20]). Secondly, small cells deployment allows UEs to com- To understand the need of an edge computing, consider
municate using mmWave in order to reduce transmission real-time packet delivery among self-driving cars that
range and interference. Thirdly, massive MIMO (multiple- requires an end-to-end delay of less than 10 ms [24].
input multiple-output) allows base stations (BSs) to use The minimum end-to-end delay for an access to cloud is
a large number of antennas (e.g., up to 16 antennas per greater than 80 ms [25], [26], which is intolerable. Edge
ii VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

increased to prevent bottleneck; and b) the traffic amount

in the core network is reduced.

Thirdly, high data rate is necessary to transmit the mas-

sive amount of data generated by a diverse range of
applications (e.g., virtual reality and remote surgery) to
edge clouds [31]. Edge servers, which can be embedded
in the BSs, allow easy access to edge clouds without the
need to access the core network. The use of mmWave
frequency bands in a small cell provides a high data rate
(a) Cloud computing. transmission.

Fourthly, high availability ensures the availability of the

cloud services at the edge. Since edge computing pushes
data and application logic to the edge clouds, the avail-
ability of the edge clouds is important.

E. APPLICATIONS OF EDGE COMPUTING IN 5G

Many applications of 5G are relying on edge computing

for real-time interaction, local processing, high data rate,
and high availability, including:

(b) Edge computing. • Healthcare, such as remote surgery and diagnostics,

FIGURE 1: Cloud computing and edge computing mod- as well as monitoring of patient vital signs and
els. data. Doctors can use a remote platform to operate
surgical tools in order to save life from a distance
where they feel safe and comfortable.
computing fulfills the sub-millisecond requirement of 5G • Entertainment and multimedia applications, such as
applications, and reduces energy consumption by around streaming HDTV or 3D TV.
30% to 40% [27], which attributes to up to five times lesser
energy consumption as compared to accessing the cloud • Virtual reality, augmented reality, and mixed reality,
[28]. such as streaming video contents to virtual reality
glasses. The size of the glasses can be reduced by
offloading computation from the glasses to edge
D. KEY REQUIREMENTS OF EDGE COMPUTING IN 5G servers.
There are four key requirements for the successful de- • Tactile internet, which is the next evolution of Inter-
ployment and operation of edge computing in 5G. While net of things, provides an ultra-responsive and ultra-
all the four key requirements are important, achieving reliable network connectivity to ensure successful
a balanced trade-off among them must be considered delivery of real-time control messages and physical
depending on the applications. tactile experiences remotely [32], [33] .
Firstly, real-time interaction, which is the fundamental • URLLC, which ensures high reliability between UEs
motivation for the use of edge computing over cloud specifically in M2M communications, supports low-
computing, ensures low latency to support delay-sensitive latency transmissions of small payloads with very
applications and services (e.g. remote surgery, tactile in- high reliability from a limited set of UEs, such as
ternet, URLLC, unmanned vehicles [29], [30] and vehicle fire alarms [34].
accident prevention) in order to improve QoS. A diverse
range of services, including decision making and data • Internet of things, such as smart appliances that
analysis, can be provided by edge servers in a real-time connect devices (e.g., household appliances) to the
manner. internet.

Secondly, local processing is feasible since data and user • Factories of the future, such as smart machines, to
requests can be processed by edge servers, rather than the improve safety and productivity. Operators can use
cloud. This means that, by reducing the traffic amount a remote platform to operate heavy machines, par-
across the connection between a small cell and the core ticularly those located at hard-to-reach and unsafe
network: a) the bandwidth of the connection can be places, from a safe and comfortable place.
VOLUME 4, 2016 iii

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

• Emergency response, whereby different kinds of data attracting potential customers and enhancing their
and information about an event or incident are QoE.
gathered from different sources at different times.
The partially available data and information are O.3 Predicting network demand to estimate the required
used to make critical decisions, and they provide a network resources to cater for the network (or user)
more complete picture of the event as time goes by. demand in a local proximity, and subsequently to
Decisions made are shared with emergency response provide optimal resource allocation to handle the
team (e.g., firefighters) in real time, even prior to local network demand. An accurate prediction of
their arrivals at the location of the event. network demand helps to decide whether a network
demand should be handled locally at the edge or at
• Intelligent transportation system, whereby drivers the cloud, and so it provides an efficient allocation
can share or gather information from traffic infor- of resources (e.g., bandwidth).
mation centers to avoid vehicles that are in danger,
or stop abruptly, in a real-time manner in order O.4 Managing location awareness to enable the geo-
to avoid accidents. In addition, unmanned vehicles graphically distributed edge servers to infer their own
can sense their surroundings and move safely in an locations and track the location of UEs to support
autonomous manner. location-based services. This enables location-based
service providers to outsource services and data to
edge clouds. For instance, mobile UEs can query and
III. TAXONOMY search for information about points of interest in
local proximities given their geographical locations.
Fig. 2 shows a taxonomy of edge computing in 5G,
The number of queries can be high, such as queries
covering objectives, computational platforms, attributes,
related to hospitals and medical advices during emer-
the use of 5G functions, performance measures, and the
gency response.
roles of edge computing in 5G.
O.5 Improving resource management to optimize net-
A. OBJECTIVES work resource utilization for network performance
enhancement due to the limited network resources
There are five main objectives of edge computing in 5G available in the edge cloud as compared to the cloud.
as follows: This is challenging as it is a multi-objective function
that must cater to a diverse range of applications, as
O.1 Improving data management to handle a large
well as user requirements and demands, which vary
amount of delay-sensitive data, which are generated
as time goes by.
by UEs, that needs to be handled locally in a real-
time manner. For instance, the local UEs in a smart
factory is expected to generate up to 1 petabyte of B. COMPUTATIONAL PLATFORMS
data daily [35]. Since accessing to cloud incurs high
latency [36], the data can be handled locally by edge Different computational platforms provide varying com-
servers. Such efficient data management is needed puting capabilities (e.g., in terms of processing loads)
to support local functions (e.g., D2D) and real-time with different characteristics (e.g., in terms of availability,
applications (e.g., remote surgery). the proximity from UEs, and the complexity of network
infrastructure) to process data at different geographical
O.2 Improving QoS to meet a diverse range of stringent locations. The computational platform can be used either
QoS requirements in order to improve quality of individually or in combination based on the network
experience (QoE) [37]. This helps to support next scenarios and application/ service requirements. As an
generation applications, including highly interactive example for the computational platform used individu-
applications and on-demand services. For instance, ally, applications and services with strict QoS require-
over-the-top (OTT) services enable online delivery ments can use edge servers to process real-time data.
of multimedia contents, which generally require low As an example for the computational platform used in
latency and high bandwidth, without the service combination, healthcare applications and services with
providers being actively involved in the control and both real-time and non-real-time data can use edge
distribution of the content [38]. This service can servers to process real-time and lightweight data, and
promote new and personalized applications that al- cloud to process heavyweight data. There are three main
low service providers to customize QoS [39]. Service computational platforms in 5G as follows:
providers must have a holistic view of subscribers
and customers, covering their contextual informa- C.1 Cloud computing gathers, processes, and stores a
tion, such as their preferences and interests. Sub- massive amount of network-wide data and infor-
sequently, the information can be personalized for mation from UEs in the network. Subsequently, it
iv VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

sends back the data and information, or decisions, delay-sensitive applications. Nevertheless, the hybrid
back to the UEs. While cloud can empower UEs with platform is more complex compared to the separate
low computational and storage capabilities, it is not cloud computing and edge computing platforms.
suitable to provide real-time services because cloud
may be far away from UEs.
C. ATTRIBUTES
C.2 Edge computing gathers, processes, and stores a mas-
sive amount of local data and information from UEs
Edge computing has three main attributes as follows:
in a local area. Edge computing has close proximity
to UEs, while cloud may be far away from UEs.
T.1 Low latency and close proximity enables edge com-
Table 1 presents a comparison of three types of edge
puting to reduce the response delay (or round-trip
computing platforms as follows:
time) suffered by UEs while accessing the traditional
C.2.1 Fog computing deploys local fog nodes which cloud. There are three main components in a re-
are local hardware devices, such as switches sponse delay: a) communication delay that depends
and routers, to provide local computation. on data rate; b) computational delay that depends
According to the OpenFog Consortium, fog on computational time; and c) propagation delay
computing is “a system-level horizontal archi- that depends on propagation distance. In general,
tecture that distributes resources and services in cloud computing, the end-to-end delay is greater
of computing, storage, control, and networking than 80ms (or 160ms for response delay) [25]. This is
anywhere along the continuum from cloud not suitable for delay-sensitive applications, such as
to things” [40]. Fog computing shares similar remote surgery and VR, that require tactile speed with
benefits to other edge computing variants (e.g., a response delay of at most 1ms [41]. In edge com-
MEC) to provide low latency and real-time puting, UEs experience reduced overall end-to-end
analytics; however, it has low storage capacity. delay and response delay due to their close proximity
to edge servers. The strategic location of edge cloud
C.2.2 MEC provides storage and computational ca-
reduces the communication and propagation delays.
pacities at the edge of the network, such as
For instance, the propagation distance is reduced to
the radio access networks (RANs) and BSs, to
tens of meters via D2D communication and in small
improve context awareness and reduce latency.
cells, and it is generally limited within a kilometer
The MEC servers, which are usually co-located
from the UEs [42].
with multiple hosts (e.g., BSs), use a virtualized
interface to access storage and computation T.2 Location awareness enables edge servers to collect
facilities. A MEC orchestrator overlooks the and process data generated by UEs on the basis of the
MEC hosts by gathering and providing real- geographical location of UEs. This allows location-
time information regarding the services offered based and personalized service provisioning to UEs,
by each host, the available resources (e.g., net- whereby edge servers can gather data generated by
work capacity and load), the network topology sources in its proximity without sending it to the
(e.g., UEs connected to the servers including cloud.
their location and networking information), as
well as managing MEC applications. T.3 Network context awareness enables edge servers to
C.1 Hybrid combines cloud computing and edge com- acquire network context information. This is because
puting so that they can cooperate. For instance, edge servers tend to possess network context in-
edge computing processes real-time data and makes formation, particularly the real-time network condi-
real-time decisions, while cloud computing processes tions (e.g., traffic load in a network cell, and radio
non-real-time data and makes non-real-time deci- access network information) and UEs’ information
sions. The hybrid infrastructure combines the ad- (e.g., allocated bandwidth and user location). The
vantages of both edge computing (i.e., real-time information allows edge servers to adapt and respond
responses) and cloud computing (i.e., high computa- to the varying network conditions and UEs, and sub-
tional and storage capabilities). Computation can be sequently to optimize network resource utilization.
performed in different layers, particularly the cloud This helps edge servers to handle a massive amount
(or the upper) layer and the fog or the edge (or the of traffic in order to improve network performance.
bottom) layer. In general, real-time tasks are executed Fine-granular information (e.g., precise individual re-
in the fog layer, and tasks requiring high computation source reservation information) can also be used to
are executed in the cloud layer. Compared to the tra- provide specific services to traffic flows in order to
ditional cloud, edge computing increases throughput cater for individual user requirements.
and reduces latency, which are important to support
VOLUME 4, 2016 v

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

TABLE 1: Comparison between fog and MEC.

Characteristics Fog MEC
Device (examples) UE, edge-cloud, cloud BS
server
Number of hops Single or multiple hops Single
Context awareness Medium High
Access technologies (ex- Wi-Fi, bluetooth, mobile Mobile networks
amples) networks
Inter-node communica- Supported Partial
tion

D. USE OF 5G FUNCTIONS throughput, energy efficiency, and spectrum utiliza-

tion [50], [51]. UEs can offload tasks and compu-
Five major enablers of edge computing in 5G are as tations to edge servers in order to empower UEs
follows: with computational capabilities. This feature of edge
U.1 Software-defined network (SDN) defines a network computing ensure successful D2D communication
architecture that separates a network into control [52], [53].
and data planes to provide flexible and agile net-
works, which helps to simplify network management
and deploy new services [43]. In general, the control E. ROLES OF EDGE COMPUTING IN 5G
plane handles policy on cloud, while the data plane
forwards traffic according to decisions made by the There are six main roles of edge computing to support
control plane. Network functions (e.g., routing) that real-time and interactive applications and services as
require real-time response can be handled by edge follows:
servers [44].
R.1 Local storage. Edge computing offloads a massive
U.2 Network function virtualization (NFV) performs net- amount of data from UEs to edge clouds. While
work functions (also known as virtual functions) in edge servers offer distributed local storage for a
virtual machines on servers, which can handle a significant amount of data, yet their storage is much
massive amount of data to provide flexible, auto- lower than that in the cloud, which has virtually
mated, and scalable networks [45], [46]. Network unlimited storage capacity. Examples of data being
demands can be processed either at the cloud or stored are computing strategies (e.g., computation
at the edge, which prevent all data and information offloading strategy [54]), metadata (e.g., timestamps
from being sent to the cloud. and geographical locations), and monitoring data.
The edge server provides different types of storage
U.3 Massive MIMO deploys multiple antenna elements to strategies to support different kinds of data. For
increase an antenna array at transmitter and receiver. instance, ephemeral storage provides temporary data
This is in accordance to the Shannon theorem [47] storage to a set of interconnected mobile devices [55],
in which the signal-to-noise ratio increases without [56].
the need to increase the transmission power, leading
to increased network capacity and energy efficiency. R.2 Local computation. Edge computing offloads com-
Using massive MIMO, multiple UEs can offload tasks putation and process from less complex (e.g., smart
to an edge server simultaneously to reduce latency phone) and highly complex (e.g., surgical tools and
and energy consumption [48]. smart factories) UEs to edge clouds. While traditional
cache and access technologies (e.g., IEEE 802.11) pro-
U.4 Dynamic access to radio access technologies provide
vide simple computation, the edge cloud is an intelli-
access to conventional access technologies, such as
gent computing system that provides local computa-
Wi-Fi, and new radio access technology (RAT) in
tion and data processing capabilities close to UEs in
5G, such as NR [49]. 5G NR is a new standard that
an independent and autonomous manner [57]. The
provides connection to a diverse range of devices for
outcomes of the computations and processes can be
achieving low latency and scalable networks, which
valuable inputs to other UEs, such as those in a smart
allow future extension to existing networks.
factory. The advantage is that edge clouds perform
U.5 D2D communication enables direct communication small tasks and provide real-time responses locally,
between neighboring UEs using ad-hoc links with- which help to reduce the cost and delay incurred to
out passing through BSs, which improves system send the required data to the cloud.
vi VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

R.3 Local data analysis. Edge computing processes and amount of tasks and data offloaded to the cloud,
performs critical and real-time data analysis on a and so it increases network performance (e.g., higher
massive amount of raw data gathered from different throughput and lower latency), which are important
applications in close proximity to generate valuable to delay-sensitive applications (e.g, remote surgery
information [58]. The capability to make data analy- and online gaming).
sis locally reduces the latency required to send data
P.3 Energy efficiency reduces energy consumption by pro-
to, as well as to wait for responses from, the cloud.
viding local functions (R.1)-(R.5). This reduces the
Subsequently, the outcomes of the local data analysis
amount of energy incurred to offload tasks and data
are used for decision making [59].
to the cloud (i.e., the energy incurred in communica-
R.4 Local decision making. Edge computing helps enti- tion), and so it increases network lifetime.
ties to make real-time decisions and corresponding
actions in an automated manner based on well-
IV. STATE OF THE ART
processed data [60]. The capability to make decisions
locally reduces involvement from more components The state of the art of edge computing in 5G networks is
and data or information exchange, leading to: a) presented according to the three categories, namely fog-
improved system availability, particularly the cloud; based, MEC-based, and hybrid solutions. Table 2 presents
and b) improved bandwidth availability. As an exam- a summary of qualitative comparison, which has been
ple, edge computing facilitates local decision making used in the literature [63], among the existing schemes,
by automated factories. Multiple entities can make covering their objectives, computational platforms, at-
decisions in a collaborative manner. tributes, performance measures, and roles.

R.5 Local operation. Edge computing enables remote

control and monitoring – particularly critical devices A. FOG BASED SOLUTIONS
including those under unsafe environment – from a
In [65], a cross-layer resource management scheme is
distance, or a more comfortable or safer place [61].
presented between optical network and fog computing
R.6 Local security enhancement. Edge computing serves over fiber networks in order to incorporate delay require-
as an additional layer between the cloud and con- ments to edge servers. The proposed scheme achieves the
nected devices in order to improve network security, objectives of improving QoS (O.2) and improving resource
including UEs with limited resources [62]. The edge management (O.5) by providing local computation (R.2).
clouds can serve as secured distributed platforms The proposed scheme uses the hybrid computational
that provide security credentials management, mal- platform (C.3), whereby the edge clouds perform real-
ware detection, software patches distribution, and time tasks and the cloud servers perform highly com-
trustworthy communications, to detect, validate, and putational and resource-intensive tasks. 5G function, in-
countermeasure attacks. The advantage is that, due cluding SDN (U.1), is used. The proposed scheme has
to the close proximity of edge computing, malicious the attribute of low latency and close proximity (T.1). The
entities can be quickly detected and isolated, and services for different applications are performed in three
real-time responses can be initiated to ameliorate layers (i.e., the cloud, fog, and UE layers). Highly com-
the effects of the attacks. This helps to minimize putational and resource-intensive services are executed
service disruptions. In addition, the scalability and at the centralized cloud layer and real-time services are
modularity nature, as well as the capabilities, of edge offloaded to the fog layer; while the UE layer performs
computing can facilitate the deployment of block functions locally at the UEs (or end devices), which
chain [64] among UEs with limited capabilities. have less computational power and storage due to their
limitations. The proposed scheme has shown to provide
higher QoS (P.2) (i.e., lower end-to-end delay).
F. PERFORMANCE MEASURES
In [66], a comparison of energy efficiency between cloud
There are three main performance measures as follows: computing and fog computing is made under different
modulation schemes, including 64 quadrature amplitude
P.1 Lower operational cost. Edge computing reduces the
modulation (QAM), 16 QAM, phase shift keying (PSK),
operational cost by providing local functions (R.1)-
and quadrature phase shift keying (QPSK), in 5G in
(R.5), instead of offloading (or sending) tasks and
order to propose an energy-efficient model to improve
data to the cloud. This reduces offloading overhead
average throughput and energy consumption per user.
(or reduces network resource consumption), such as
The proposed scheme achieves the objectives of improv-
bandwidth.
ing QoS (O.2) by providing local computation (R.2) and
P.2 Higher QoS. Edge computing improves QoS by pro- local operations (R.5). The proposed scheme uses edge
viding local functions (R.1)-(R.5). This reduces the computing (C.2), particularly fog computing, to analyze
VOLUME 4, 2016 vii

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

FIGURE 2: Taxonomy of edge computing in 5G.

fog computing and its energy consumption as compared node is a UE with 3Cs that manage and manipulate
to cloud computing. 5G function, including dynamic the edge node of vFogs. In this architecture, a regular
access to RATs (U.4), is used. The proposed scheme has extreme node informs its corresponding super extreme
the attribute of low latency and close proximity (T.1). node about its available resources, and then receive and
Different RATs (i.e., 3G, 4G, and 5G) serve a number of dif- execute networking tasks assigned by the super extreme
ferent UEs. An energy efficiency model is proposed based node. The proposed architecture has shown to provide
on throughput, energy consumption, and the energy lower operational cost (P.1) and higher QoS (P.2) (i.e.,
consumption level under fog environment. The proposed lower decision making delay).
scheme has shown to improve energy efficiency (P.3).

In [67], an architecture that enables edge servers to pro- B. MEC BASED SOLUTIONS
vide caching, computing, and communications functions
(also known as 3Cs) is proposed so that content and In [68], an architecture is presented to perform energy-
service providers can deploy their functions, services, aware offloading, whereby each mobile UE decides
and contents closed to UEs. The proposed architecture whether to perform or offload computational tasks to
achieves the objectives of improving QoS (O.2) and im- MEC server, in order to reduce energy consumption of
proving resource management (O.5) by providing local MEC. The UEs are heterogeneous in nature as they have
storage (R.1) and local computation (R.2). The proposed different communication and computing capabilities, and
architecture uses edge computing (C.2), particularly fog the energy consumption of the computational tasks at
computing, to reduce processing delay. 5G functions, in- the mobile UEs is higher than that in the MEC server
cluding SDN (U.1) and NFV (U.2), are used. The proposed [69]. The proposed architecture achieves the objectives
architecture has the attributes of low latency and close of improving data management (O.1) and improving QoS
proximity (T.1) and network context awareness (T.3) to (O.2) by providing local computation (R.2) and local de-
acquire network information and traffic distribution. The cision making (R.4). The proposed architecture uses edge
architecture consists of: a) virtual fog (vFog) which is computing (C.2) to perform energy-aware offloading. 5G
a framework that empowers UEs with 3Cs using NFV function, including dynamic access to RATs (U.4), is used.
so that service provisioning becomes flexible; b) hyper The proposed architecture has the attribute of low latency
fog which is a constellation of vFogs that allows data and close proximity (T.1). There are three main steps.
exchange and processing among the vFogs in order to Firstly, mobile UEs are classified according to their energy
provide resources from more than a single vFog; c) consumption in computation and file transmission, and
regular extreme node, which is a UE with processing the transmission delay between the mobile UEs and the
and communication capabilities; and d) super extreme MEC. There are three main categories: a) type 1 UEs
viii VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

use MEC server for computation; b) type 2 UEs perform context awareness (T.3) to acquire network information
computation themselves; and c) type 3 UEs can choose to and traffic distribution. The UEs are categorized based on
perform computation either at MEC servers or by them- their computational capacity and links (i.e., cellular and
selves. Secondly, priorities are given to the different UEs D2D links). The UEs use graph matching to determine
based on their energy consumption, as well as available whether to perform tasks locally or to offload them to the
channels and their channel quality. In general, type 1 UEs edge nodes via D2D in order to achieve energy efficiency.
enjoy higher priorities due to their limited computational The graph matching algorithm, which represents nodes
capabilities and the need to offload computational tasks and links in a graph, has two main stages: a) prunes
to the MEC server in order to satisfy the delay constraint. out nodes without tasks; and b) creates a sub-graph that
Thirdly, channels are allocated for UEs with different consists of nodes with tasks (also called task nodes),
priorities. Since UEs with higher priorities are offloaded, replicates them, and connects them to edge nodes of the
there are lower number of UEs competing for channels. graph. A task node performs a task locally if it can be
The proposed architecture has shown to provide higher matched by its own replica, and it offloads the task to
energy efficiency (P.3). another node if it can be matched with the other node.
The proposed architecture has shown to provide higher
In [12], MEC services are autonomously created by the
energy efficiency (P.3).
nearest edge server in order to provide mobile UEs
with seamless QoE in video streaming. The proposed In [70], a predictive and proactive caching approach is
scheme achieves the objectives of improving QoS (O.2) introduced in order to reduce peak traffic demands. The
and predicting network demand (O.3) by providing local proposed scheme achieves the objectives of improving
storage (R.1), local computation (R.2), and local decision data management (O.1) and predicting network demand
making (R.4). The proposed scheme uses edge comput- (O.3) by providing local storage (R.1) and local computa-
ing (C.2) to perform uninterrupted video streaming. 5G tion (R.2). The proposed scheme uses edge computing
function, including D2D communication (U.5), is used. (C.2) to perform proactive caching at the edge of the
The proposed scheme has the attributes of low latency network or at UEs. 5G function, including D2D com-
and close proximity (T.1), location awareness (T.2), and munication (U.5), is used. The proposed scheme has
network context awareness (T.3). The edge server receives the attributes of low latency and close proximity (T.1),
all or part of a content (e.g., video) from the cloud, so and network context awareness (T.3) to acquire network
that the content can be transmitted to UEs with reduced information and traffic distribution. Popular contents are
delay. Hence, the quality of the content is good as long cached in edge servers, BSs, or UEs during off peak
as a UE is in the vicinity of the edge server. In general, times.WThe popularity of a content is based on the UEs’
the UEs receive contents from the edge server to reduce behavior and the frequency of the BS requesting for the
delay (and hence, higher quality streaming); however, content. When a BS requests for a particular content,
if the contents are unavailable in the edge server, the there are two possibilities: a) the content is available at
UEs would receive contents from the cloud, which in- an influential UE, who had possessed or processed the
creases delay (and hence, lower quality streaming). There content in the past, and so the content is delivered from
are two main mechanisms to ensure seamless content the influential UE to the BS via D2D; and b) the content is
transmission. Firstly, migration enables seamless content unavailable at any influential UEs, and so the content is
transmission when a UE moves from the vicinity of delivered from the core network to the BS. The proposed
an edge server to another. Secondly, handover enables scheme has shown to provide lower operational cost (P.1).
seamless content transmission when a UE handover from
In [71], an application-aware traffic redirection mecha-
a network provider to another, which reduces delay (and
nism is proposed for MEC in order to reduce response
hence, higher quality streaming). The proposed scheme
time and bandwidth consumption. The proposed scheme
has shown to provide higher QoS (P.2) (i.e., lower end-to-
achieves the objectives of improving data management
end delay).
(O.1) and improving QoS (O.2) by providing local com-
In [52], a D2D architecture is proposed for a mas- putation (R.2), local decision making (R.4), and local
sive number of UEs to execute collaborative tasks in operation (R.5). The proposed scheme uses edge com-
an energy-efficient manner. The proposed architecture puting (C.2). 5G functions, including dynamic access to
achieves the objectives of improving QoS (O.2) and im- RATs (U.4) and D2D communication (U.5), are used. The
proving resource management (O.5) by providing local proposed scheme has the attributes of low latency and
computation (R.2), local decision making (R.4), and local close proximity (T.1) and network context awareness (T.3)
operation (R.5). The proposed architecture uses edge to acquire network information and traffic distribution
computing (C.2) to perform energy-efficient task offload- (T.3). The MEC controller allows UEs to offload (or
ing. 5G function, including D2D communication (U.5), redirect) the traffic of an application to MEC at the edge
is used. The proposed architecture has the attributes of the network when the bandwidth requirement of the
of low latency and close proximity (T.1), and network traffic exceeds a preset threshold. Subsequently, the UEs
VOLUME 4, 2016 ix

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

can access the application and its traffic. The proposed In [75], a group of vehicular neighboring nodes (or VNG)
scheme has shown to provide lower operational cost (P.1) is dynamically managed using SDN to improve control
and higher QoS (P.2) (i.e., lower response time). over network and its resources in vehicular networks.
The proposed scheme achieves the objectives of im-
In [72], a virtualized multi-access edge computing frame- proving data management (O.1), improving QoS (O.2),
work is proposed to increase available bandwidth and and improving resource management (O.5) by providing
reduce end-to-end delay in an intelligent manner in local computation (R.2), local decision making (R.4), and
Internet of things. The proposed framework achieves the local operation (R.5). The proposed scheme uses edge
objectives of improving data management (O.1) and im- computing (C.2). 5G function, including SDN (U.1), is
proving QoS (O.2) by providing local computation (R.2), used. The proposed architecture has the attributes of low
and local decision making (R.4). The proposed framework latency and close proximity (T.1), and network context
uses edge computing (C.2). 5G function, including NFV information and traffic distribution (T.3). The proposed
(U.2), is used. The proposed framework has the attributes scheme integrates SDN to MEC in order to strengthen
of low latency and close proximity (T.1), and network network control (e.g., a unified network control of het-
context awareness (T.3) to acquire network information erogeneous networks) at the edge of the network for
and traffic distribution. The proposed framework uses achieving a flexible network control and management.
MEC to perform virtualized multi-access computing at Real-time instructions (e.g., safety messages) is passed
the edge of the network. Hardware devices are disaggre- from road side units to vehicles in order to monitor
gated and virtualized into layers that provide different network states (i.e., the available resources of vehicles)
control functions (e.g., traffic offloading), services (e.g., in order to make effective decisions (i.e., road blocks and
computational and storage capabilities), and resources route changes). Using SDN, the edge of the network is
(e.g., computing and storage resources) using NFV. In segregated into three layers: a) the control plane enables
addition, traffic offloading provides network traffic in- the MEC to obtain the global knowledge of network
formation, such as the number of packets, as well as states for making optimal decisions (i.e., network-level
the priority level and type of traffic, based on data flow. decisions for efficient networking and fault diagnosis)
Traffic is prioritized and segregated into three categories, with lower response time; b) the social plane, which is
namely high-, medium-, and low-priority traffic, based abstracted for communication among VNGs, enables the
on the packet flow rate and type of traffic, as well as SDN switch to separate and forward sociality flows, each
the number of packets in the queue. Low-priority packets of which consists of data packets that indicate the key
are dropped when signal strength is low and congestion features of a VNG (e.g., the strength of a relationship,
occurs. The proposed framework has shown to provide contact time, contact frequency, and the contact method)
lower operational cost (P.1) and higher QoS (P.2) (i.e., among vehicles so that suitable vehicles can be selected
lower end-to-end delay). to form strong and weak ties. As an example, two work-
mates from the same office leaving a parking area on a
In [73], a fiber wireless (FiWi) access architecture is daily basis can form a strong tie. As another example,
introduced to improve MEC services (e.g., traffic and random vehicles on the road can form temporary weak
network performance monitoring). The proposed archi- ties; and c) the data plane provides data transmission.
tecture achieves the objectives of improving resource The proposed scheme has shown to provide higher QoS
management (O.5) by providing local computation (R.2), (P.2) (i.e., lower end-to-end delay).
local decision making (R.4), and local operation (R.5).
The proposed scheme uses edge computing (C.2). 5G In [76], a non-standalone (i.e., disconnected from the In-
functions, including dynamic access to RATs (U.4) and ternet) MEC-based architecture is presented for mission-
D2D communication (U.5), are used. The proposed ar- critical public safety services in order to achieve the
chitecture has the attributes of low latency and close delay requirement (i.e., less than 1 ms (ideal) or 10 ms
proximity (T.1) and network context awareness (T.3) to (maximum) of round trip time) of 5G. The proposed
acquire network information and traffic distribution. In architecture achieves the objective of improving QoS
the proposed architecture, BSs serve as rational service (O.2) by providing local computation (R.2), local decision
centers that provide updated information (i.e., traffic de- making (R.4), and local operation (R.5). The proposed
mand and RAT) to backhaul in order to provide intelligent architecture uses edge computing (C.2). 5G functions,
and energy-efficient schemes. MEC is operating over FiWi including SDN (U.1), NFV (U.2), and dynamic access to
[74], and ethernet is used to transfer traffic from RAN. The RATs (U.4), are used. The proposed architecture has the
FiWi, along with ethernet, provides a framework for back- attribute of low latency and close proximity (T.1). MEC is
haul and broadband access. The proposed architecture used to provide a flexible architecture, whereby the user
has shown to provide higher QoS (P.2) (i.e., lower queuing plane, which is the bottom layer, consists of UEs that
delay in the data buffer) and lower energy consumption can be grouped into virtual groups (or clusters) based
(P.3). on their ownership, as well as co-location and co-service
x VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

relationships that define the relative location between (R.4), and local operation (R.5). The proposed archi-
a cluster and a service requested by the UEs. MEC is tecture uses hybrid computing (C.3) to optimally dis-
deployed close to the UEs, and the flexibility of the tribute tasks among cloud, MEC, and mobile UEs. 5G
architecture allows the location and structure of the MEC function, including dynamic access to RATs (U.4), is
to be customized and redefined as time goes by. The com- used. The proposed architecture has the attributes of low
putational resources (e.g., servers, processors, and cloud), latency and close proximity (T.1) and network context
as well as radio interfaces and schemes (e.g., modulation awareness (T.3). Tasks are split and offloaded among
schemes and TDMA) are distributed in different slices in cloud, MEC, and mobile UEs based on the task require-
network slicing, which enables virtualization by running ments: a) UEs process tasks that require less processing
multiple logical networks on a shared physical network and computational capabilities; b) MEC server processes
infrastructure. The key benefit of network slicing is that delay-sensitive tasks; and c) cloud processes non-delay-
it provides an end-to-end virtual network encompassing sensitive tasks. Both MEC and cloud process tasks that
networking, computation, and storage functions. Urgent require higher processing and computational capabilities.
services (e.g., mission critical services) are executed in The proposed architecture has shown to provide lower
higher priority slices (e.g., real-time services such as life- operational cost (P.1) and higher QoS (P.2).
saving services in e-health). Hence, additional resources
In [79], a real-time, context-aware, service-composition,
are allocated to higher priority slices to serve the urgent
and collaborative architecture is proposed to deliver fast
services. The proposed architecture has shown to provide
composite service, which is the consolidation of multiple
higher QoS (P.2) (i.e., lower end-to-end delay).
services supported by the collaboration of different hard-
ware (e.g., UEs, edge clouds, and cloud) and software with
C. HYBRID SOLUTIONS different capabilities. The proposed architecture achieves
the objectives of improving data management (O.1) and
In [77], a D2D-based mobile edge and fog computing improving QoS (O.2) by providing local computation
architecture is introduced to enable collaborative com- (R.2), local data analysis (R.3), local decision making (R.4),
puting, which performs tasks in more than a single and local operation (R.5). The proposed architecture uses
computing platforms or paradigms, in order to enhance hybrid computing (C.3) that enables collaboration among
MEC. The proposed architecture achieves the objectives cloud, MEC, and UEs. 5G function, including dynamic
of improving data management (O.1) and improving QoS access to RATs (U.4), is used. The proposed architecture
(O.2) by providing local storage (R.1), local computation has the attributes of low latency and close proximity (T.1)
(R.2), local data analysis (R.3), and local decision making and network context awareness (T.3). Frequently accessed
(R.4). The proposed architecture uses hybrid computing blocks, which are small units decomposed from a file,
(C.3) to exploit D2D communication in collaborative are stored (or cached) in MEC servers. Blocks requested
environment. 5G function, including D2D communica- by more than a single server are replicated and cached
tion (U.5), is used. The proposed architecture has the in other MEC servers based on file types and contents.
attribute of low latency and close proximity (T.1). Each UE This helps to reduce the end-to-end delay incurred to
initiates a service request and send it to the nearest relay access the cloud. The proposed architecture has shown
gateway, which has connection to the core network (or to provide lower operational cost (P.1) and higher QoS
cloud). The service handler of a relay gateway, which has (P.2).
information about the available services, decides whether
the requested service should be performed locally or
forwarded to another relay gateway that can perform the V. OPEN RESEARCH ISSUES
service. The decision is based on the availability of the This section highlights the open research issues for a suc-
service (e.g., the processing, computational, and storage cessful deployment of edge cloud in the 5G environment.
capabilities, as well as delay requirements) at the relay
gateway and its neighboring gateways. The proposed
architecture has shown to provide lower operational cost A. SERVICE ENHANCEMENT: QOE
(P.1) and higher QoS (P.2) (i.e., lower end-to-end and
round-trip delays). QoE is a measure of the overall customer satisfaction
level with a service provider. QoE is related to, but differs
In [78], a context-aware, real-time collaborative archi- from, QoS, which embodies the notion that the hardware
tecture is proposed to manage heterogeneous resources characteristics (e.g., the storage capacity and the number
(e.g., different storage and computational capabilities in of processors in the servers [80]) and software character-
different computational platforms/ layers) at the edge istics (e.g., the interface development) can be measured,
of the network. The proposed architecture achieves the improved, and guaranteed. The challenge is to achieve
objective of improving resource management (O.5) by a balanced trade-off between: a) higher availability or
providing local computation (R.2), local decision making seamless connectivity of an application, which can be
VOLUME 4, 2016 xi

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 VOLUME 4, 2016
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
TABLE 2: SUMMARY OF OBJECTIVES , CHALLENGES , METRICS , CHARACTERISTICS , AND PERFORMANCE MEASURES OF CLUSTERING SCHEMES FOR 5G NETWORKS
Reference Objectives Computing Edge Attributes 5G functions Performance Roles of
computing Edge computing
T.3 Network context information

U.1 Software defined networks

U.2 Network function virtualization

T.2 Location awareness
E.1 MEC

U.3 Massive MIMO

U.5 D2D communication

P.1 operational cost

R.2 Local computation

R.5 Local operation

P.2 QoS
O.1 Improving data management

O.2 Enhancing QoS

O.3 Predicting network demand

O.4 Managing location awareness

C.1 Cloud computing

C.2 Edge computing

C.3 Hybrid computing

T.1 Low latency

P.3 Energy efficiency

R.1 Local storage

R.3 Local data analysis

R.4 Local decision making

R.6 Local security enhancement

O.5 Managing resource

E.2 Fog Computing

U.4 Dynamic access to RAT

10.1109/ACCESS.2019.2938534, IEEE Access

Guo et al [65] X X X X X X X X
Kitanov et al [66] X X X X X X X X
Markakis et al [67] X X X X X X X X X X X X
Zhang et al [68] X X X X X X X X X
Taleb et al [12] X X X X X X X X X X X X
Chen et al [52] X X X X X X X X X X
Bastug et al [70] X X X X X X X X X X
Huang et al [71] X X X X X X X X X X X X X
Hsieh et al [72] X X X X X X X X X X X
Rimal et al [73] X X X X X X X X X X X X
Huang et al [75] X X X X X X X X X X X X
Solozabal et al [76] X X X X X X X X X X X
Singh et al [77] X X X X X X X X X X X X
Tran et al [78] X X X X X X X X X X X
Ridhawi et al [79] X X X X X X X X X X X X

xii
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

provided by the cloud when a UE is out of the vicinity proposed solutions enable heterogeneous UEs to com-
of the edge server; and b) higher QoE of the application, municate with edge servers.
which can be provided by the edge cloud when the UEs
are in the vicinity of the edge server, in order to re-
D. SECURITY AND PRIVACY
duce delay and jitter. Hence, collaborative computational
approaches, such as hybrid computing (C.3), can be While security and privacy is enhanced in edge comput-
used. Edge computing can be used to maintain network ing as data do not travel across a network, there are two
or service states (e.g. the availability and cost of the main problems that can increase network vulnerability
links, as well as the way a switch forwards the traffic) at the edge of network. Firstly, the dynamic environment
for evolving applications (e.g., 4K video streaming [81]) causes the data and network requirements of different
and offer proxying functionality on behalf of UEs. By network entities to vary rapidly. Secondly, the increasing
maintaining the network states, the trade-off between number of devices communicating with each other must
the availability and QoE performance can be achieved require a scalable solution. Hence, trust and security
with reduced signaling overhead incurred by network management must address the aforementioned problems
processes (e.g., handover). The signaling messages can in order to address network vulnerability; however, this
also be aggregated to reduce signaling overhead. This may incur high complexity and cost. Enhancing security
leads to reduced network congestion, hence improving and privacy is significant due to the importance of the
network scalability and network performance (e.g., higher data (e.g., health information). There are two potential
throughput [82]). Addressing this open issue can provide solutions. Firstly, applications running on edge cloud
improvement in QoS (P.2). must be blind/ unaware to the raw information (or
unprocessed data). So, the raw information (e.g., per-
sonal data including healthcare information) must be
B. STANDARDIZATION OF PROTOCOLS
encrypted or processed. Secondly, raw information can
Standardization of protocols requires standardizing bod- be removed prior to reaching the edge cloud to ensure
ies or organizations to provide a set of universally ac- privacy [84], [85].
ceptable rules for edge computing in 5G environment.
There are two main challenges. Firstly, it is difficult to VI. CONCLUSION
agree upon a standard (e.g., the location and capabilities
of the edge cloud) due to its flexibility and diversified In this paper, we present a review of the state-of-the
customization by different vendors. Secondly, a large art development in edge computing, including fog-based,
number of heterogeneous UEs use different interfaces to MEC-based, and hybrid solutions, in 5G networks. A
communicate with the edge cloud. Standardization effort, taxonomy is established in which the edge computing
such as the initiative from the European Telecommu- approaches are classified according to different charac-
nications Standards Institute (ETSI) [83], has been put teristics (e.g, objectives, computational platforms, and
in place so that heterogeneous UEs can communicate attributes) and the features of edge computing are pre-
with edge servers, and different layers and computation sented. The key requirements of edge computing are
paradigms can collaborate among themselves, in a multi- to provide real-time interaction, local processing, high
vendor environment. data rate, and high availability. Edge computing improves
network performance to support and deploy different
scenarios, such as remote surgery. Open issues for the
C. ADDRESSING HETEROGENEITY successful deployment of edge computing in 5G are iden-
tified, including service enhancement, standardization,
Heterogeneity in communication (e.g., transmission as well as addressing heterogeneity and security vulner-
range and data rate) and computing (e.g., hardware abilities. A qualitative comparison among the existing
architecture and operating systems) technologies in edge schemes in the literature is presented, and it shows
computing for 5G has resulted in difficulties in develop- the research gaps in this topic whereby the missing
ing a solution that is portable across different environ- ticks represent potential open issues that can be further
ment. Software-based (or programming-based) schemes investigated. Although the deployment of edge computing
may develop a programming-model for edge nodes to in 5G provides numerous benefits, the convergence of
facilitate the execution of workloads simultaneously at both edge computing and 5G brings about new issues
multiple hardware levels [2]. However, a comprehensive that should be resolved in the near future.
distributed computing system must allow the different
schemes to operate in a collaborative manner. Data and
task-level parallelism splits workload into independent REFERENCES
and smaller tasks that can be executed in parallel across [1] W. Z. Khan, E. Ahmed, S. Hakak, I. Yaqoob, and A. Ahmed, “Edge com-
different hardware and layers in edge clouds [12]. The puting: A survey,” Future Generation Computer Systems, vol. 97, pp.

VOLUME 4, 2016 xiii

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

219–235, Aug 2019. [20] Z. Pi, J. Choi, and R. Heath, “Millimeter-wave gigabit broadband evo-
lution toward 5G: fixed access and backhaul,” IEEE Communications
[2] N. Hassan, S. Gillani, E. Ahmed, I. Yaqoob, and M. Imran, “The role of Magazine, vol. 54, no. 4, pp. 138–144, Apr 2016.
edge computing in internet of things,” IEEE Communications Magazine,
vol. 56, no. 11, pp. 110–115, Nov 2018. [21] A.Ateya, A.Muthanna, I.Gudkova, A.Abuarqoub, A.Vybornova, and
A.Koucheryavy. “Development of intelligent core network for tactile
[3] E.Ahmed, A.Ahmed, I.Yaqoob, J.Shuja, A.Gani, M. Imran, and Muham- internet and future smart systems.” Journal of Sensor and Actuator
mad Shoaib. “ Bringing computation closer toward the user network: Networks, vol. 7, no. 1, Jan 2018.
Is edge computing the solution?”, IEEE Communications Magazine,
55(11):138–144, nov 2017. [22] Z.Ning, X.Kong, F.Xia, W.Hou, and X.Wang, "Green and sustainable cloud
of things: Enabling collaborative edge computing", IEEE Communica-
[4] J.Shuja, S.Mustafa, R. W.Ahmad, S .A. Madani, A.Gani, and M. K.Khan,
tions Magazine, vol. 57, no. 1, pp.72-78, 2018.
“Analysis of vector code offloading framework in heterogeneous cloud
and edge architectures”, IEEE Access, vol 5, pp. 24542–24554, 2017. [23] K.Yeow, A.Gani, R.W.Ahmad, J.Rodrigues, K.Ko, "Decentralized consen-
sus for edge-centric internet of things: A review, taxonomy, and research
[5] P.Popovski, K.Floe Trillingsgaard, O.Simeone, and G.Durisi. “5G wire-
issues", IEEE Access, vol. 6, no. 6, pp. 1513-1524, Dec 2017.
less network slicing for eMBB, URLLC, and mMTC: A communication-
theoretic view.” IEEE Access, vol. 6, pp. 55765–55779, 2018.
[24] M. Simsek, A. Aijaz, M. Dohler, J. Sachs, and G. Fettweis, “5G-enabled
[6] A.Anand, G D.Veciana, and S.Shakkottai. , “Joint scheduling of URLLC tactile internet,” IEEE Journal on Selected Areas in Communications,
and eMBB traffic in 5g wireless networks.” ,In IEEE INFOCOM 2018 - vol. 34, no. 3, pp. 460–473, Mar 2016.
IEEE Conference on Computer Communications. IEEE, apr 2018.
[25] S. Choy, B. Wong, G. Simon, and C. Rosenberg, “The brewing storm
[7] J. M. Hamamreh, E.Basar, and H.Arslan , “OFDM-subcarrier index selec- in cloud gaming: A measurement study on cloud to end-user latency,”
tion for enhancing security and reliability of 5g URLLC services.” IEEE in 2012 11th Annual Workshop on Network and Systems Support for
Access, vol. 5, pp. 25863–25875, 2017. Games (NetGames). IEEE, Nov 2012.

[8] P. Gallo, K. Kosek-Szott, S. Szott, and I. Tinnirello, “CADWAN: A control [26] Z. Zhao, K. Hwang, and J. Villeta, “Game cloud design with virtualized
architecture for dense WiFi access networks,” IEEE Communications CPU/GPU servers and initial performance results”, in Proceedings of the
Magazine, vol. 56, no. 1, pp. 194–201, Jan 2018. 3rd workshop on Scientific Cloud Computing Date - ScienceCloud ACM
Press, 2012.
[9] A. Ahmed and E. Ahmed, “A survey on mobile edge computing,” in
2016 10th International Conference on Intelligent Systems and Control [27] W. Shi and S. Dustdar, “The promise of edge computing,” Computer,
(ISCO). IEEE, Jan 2016. vol. 49, no. 5, pp. 78–81, May 2016.

[10] Q. V. Pham , F. Fang, V.N.Ha, M.Le, Z.Ding, L.B.Le, W.J.Hwang, "A [28] W. Hu, Y. Gao, K. Ha, J. Wang, B. Amos, Z. Chen, P. Pillai, and M. Satya-
Survey of Multi-Access Edge Computing in 5G and Beyond: Funda- narayanan, “Quantifying the impact of edge computing on mobile appli-
mentals, Technology Integration, and State-of-the-Art", arXiv preprint cations,” in Proceedings of the 7th ACM SIGOPS Asia-Pacific Workshop
arXiv:1906.08452. 2019 Jun 20. on Systems - APSys '16. ACM Press, 2016.

[11] K.Antonakoglou, X.Xu, E.Steinbach, T.Mahmoodi, and M.Dohler, “To- [29] M.Fahad, F.Aadil, Z.Rehman, S.Khan, P.Azmat, S.Khan, M.Jaime,
ward haptic communications over the 5G tactile internet.” IEEE Com- L.Haoxiang, W.Jong, W.Lee, I.Mehmood, "Grey wolf optimization based
munications Surveys & Tutorials, vol. 20, no. 4, pp. 3034–3059, 2018. clustering algorithm for vehicular ad-hoc networks.", Computers &
Electrical Engineering, vol. 70, pp. 853-870, 2018.
[12] T. Taleb, S. Dutta, A. Ksentini, M. Iqbal, and H. Flinck, “Mobile edge
computing potential in making cities smarter,” IEEE Communications [30] F Aadil, A Raza, M Khan, M Maqsood, I Mehmood, S. Rho, "Energy aware
Magazine, vol. 55, no. 3, pp. 38–43, Mar 2017. cluster-based routing in flying ad-hoc networks." Sensors, vol. 18, no. 5,
pp. 1413, 2018
[13] Y. Yu, “Mobile edge computing towards 5G: Vision, recent progress, and
open challenges,” China Communications, vol. 13, no. Supplement2, pp. [31] P. Porambage, J. Okwuibe, M. Liyanage, M. Ylianttila, and T. Taleb, “Sur-
89–99, 2016. vey on multi-access edge computing for internet of things realization,”
[14] S.Raza, S.Wang, M.Ahmed, M.R.Anwar, "A Survey on Vehicular Edge IEEE Communications Surveys & Tutorials, vol. 20, no. 4, pp. 2961–2991,
Computing: Architecture, Applications, Technical Issues, and Future Di- 2018.
rections." Wireless Communications and Mobile Computing, vol.2019,
[32] M.Shahid, A.Bo, A.Dulaimi, and K.Tsang. "Guest Editorial 5G Tactile
2019.
Internet: An Application for Industrial Automation.", IEEE Transactions
[15] S. York and R. Poynter, “Global mobile market research in 2017,” in on Industrial Informatics, vol. 15, no. 5, pp. 2992-2994, 2019: .
Mobile Research. Springer Fachmedien Wiesbaden, Aug 2017, pp. 1–
[33] A.Mohammad, K.A.Harras, and S.Zeadally. "Fog computing for 5G tactile
14.
industrial internet of things: Qoe-aware resource allocation model."
[16] Z. Cui and J. Wang, “Enhanced software-defined network controller to IEEE Transactions on Industrial Informatics, vol. 15, no.5, pp. 3085-3092,
support ad-hoc radio access networks,” Oct. 23 2018, uS Patent App. 2019.
10/111,127.
[34] Jaya Rao and Sophie Vrzic. “Packet duplication for URLLC in 5G: Archi-
[17] I. A. Ridhawi, M. Aloqaily, Y. Kotb, Y. A. Ridhawi, and Y. Jararweh, “A col- tectural enhancements and performance analysis.” IEEE Network, vol.
laborative mobile edge computing and user solution for service compo- 32, no 2, pp. 32–40, Mar 2018.
sition in 5G systems,” Transactions on Emerging Telecommunications
Technologies, vol. 29, no. 11, p. e3446, Jun 2018. [35] R. Miller, “Data center first: Intels vision for a
data-driven world”, https://datacenterfrontier.com/
[18] B. Yang, Z. Yu, J. Lan, R. Zhang, J. Zhou, and W. Hong, “Digital data-center-first-intels-vision-for-a-data-driven-world/, 2017
beamforming-based massive MIMO transceiver for 5G millimeter-wave (accessed April 24, 2019).
communications,” IEEE Transactions on Microwave Theory and Tech-
niques, vol. 66, no. 7, pp. 3403–3418, Jul 2018. [36] X. Sun and N. Ansari, “Green cloudlet network: A distributed green
mobile cloud network,” IEEE Network, vol. 31, no. 1, pp. 64–70, Jan 2017.
[19] N. C. Luong, P. Wang, D. Niyato, Y.-C. Liang, Z. Han, and F. Hou, “Applica-
tions of economic and pricing models for resource management in 5G [37] Y. C. Hu, M. Patel, D. Sabella, N. Sprecher, and V. Young, “Mobile edge
wireless networks: A survey,” IEEE Communications Surveys & Tutorials, computing—a key technology towards 5G,” ETSI white paper, vol. 11,
pp. 1–1, 2018. no. 11, pp. 1–16, 2015.

xiv VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

[38] Z. Zhao, L. Guardalben, M. Karimzadeh, J. Silva, T. Braun, and S. Sar- [57] A. Ndikumana, N. H. Tran, T. M. Ho, Z. Han, W. Saad, D. Niyato, and
gento, “Mobility prediction-assisted over-the-top edge prefetching for C. S. Hong, “Joint communication, computation, caching, and control
hierarchical VANETs,” IEEE Journal on Selected Areas in Communica- in big data multi-access edge computing,” IEEE Transactions on Mobile
tions, vol. 36, no. 8, pp. 1786–1801, Aug 2018. Computing, pp. 1–1, 2019.

[39] E. Chirivella-Perez, J. Gutiérrez-Aguado, J. M. Alcaraz-Calero, and [58] C. Li, Y. Xue, J. Wang, W. Zhang, and T. Li, “Edge-oriented computing
Q. Wang, “Nfvmon: enabling multioperator flow monitoring in 5G mo- paradigms,” ACM Computing Surveys, vol. 51, no. 2, pp. 1–34, Apr 2018.
bile edge computing,” Wireless Communications and Mobile Comput-
ing, vol. 2018, 2018. [59] E. Ahmed, I. Yaqoob, I. A. T. Hashem, I. Khan, A. I. A. Ahmed, M. Imran,
and A. V. Vasilakos, “The role of big data analytics in internet of things,”
[40] F. Bonomi, R. Milito, J. Zhu, and S. Addepalli, “Fog computing and its role Computer Networks, vol. 129, pp. 459–471, Dec 2017.
in the internet of things,” in Proceedings of the first edition of the MCC
workshop on Mobile cloud computing - MCC '12. ACM Press, 2012. [60] M. M. Hussain, M. S. Alam, and M. M. S. Beg, “Feasibility of fog com-
puting in smart grid architectures,” in Proceedings of 2nd International
[41] G. Intelligence, “Understanding 5G: Perspectives on future technologi-
Conference on Communication, Computing and Networking. Springer
cal advancements in mobile,” White paper, pp. 1–26, 2014.
Singapore, Sep 2018, pp. 999–1010.
[42] Y. Mao, C. You, J. Zhang, K. Huang, and K. B. Letaief, “A survey on mobile
edge computing: The communication perspective,” IEEE Communica- [61] C. Xu, K. Wang, P. Li, S. Guo, J. Luo, B. Ye, and M. Guo, “Making big data
tions Surveys & Tutorials, vol. 19, no. 4, pp. 2322–2358, 2017. open in edges: A resource-efficient blockchain-based approach,” IEEE
Transactions on Parallel and Distributed Systems, vol. 30, no. 4, pp. 870–
[43] Y. Li and M. Chen, “Software-defined network function virtualization: A 882, Apr 2019.
survey,” IEEE Access, vol. 3, pp. 2542–2553, 2015.
[62] M. Ali, H. S. M. Bilal, M. A. Razzaq, J. Khan, S. Lee, M. Idris, M. Aazam,
[44] Y.Jararweh, C.Mavromoustakis, D.B. Rawat, M.H.Rehmani, "SDN, NFV, T. Choi, S. C. Han, and B. H. Kang, “IoTFLiP: IoT-based flipped learn-
and Mobile Edge Computing with QoE Support for 5G", Transactions ing platform for medical education,” Digital Communications and Net-
on Emerging Telecommunications Technologies, vol. 29, no. 11, pp. works, vol. 3, no. 3, pp. 188–194, Aug 2017.
3536.Nov 2018.
[63] P. Mach, Z. Becvar, "Mobile edge computing: A survey on architecture
[45] A. J. Gonzalez, G. Nencioni, A. Kamisinski, B. E. Helvik, and P. E. Hee- and computation offloading.", IEEE Communications Surveys & Tutori-
gaard, “Dependability of the NFV orchestrator: State of the art and als, vol. 19, no. 3, pp. 1628-1656, 2017.
research challenges,” IEEE Communications Surveys & Tutorials, vol. 20,
no. 4, pp. 3307–3329, 2018. [64] R. Yang, F. R. Yu, P. Si, Z. Yang, and Y. Zhang, “Integrated blockchain
and edge computing systems: A survey, some research issues and chal-
[46] I.Sarrigiannis, E.Kartsakli, K.Ramantas, A.Antonopoulos, C.Verikoukis,
lenges,” IEEE Communications Surveys & Tutorials, pp. 1–1, 2019.
"Application and Network VNF migration in a MEC-enabled 5G Archi-
tecture", 23rd IEEE International Workshop on Computer Aided Mod- [65] S. Guo, S. Shao, Y. Wang, and H. Yang, “Cross stratum resources protec-
eling and Design of Communication Links and Networks (CAMAD), pp. tion in fog-computing-based radio over fiber networks for 5G services,”
1-6 , 2018. Optical Fiber Technology, vol. 37, pp. 61–68, Sep 2017.
[47] C. E. Shannon, “A mathematical theory of communication,” Bell System
[66] S. Kitanov and T. Janevski, “Energy efficiency of fog computing and
Technical Journal, vol. 27, no. 3, pp. 379–423, Jul 1948.
networking services in 5G networks,” in IEEE EUROCON 2017 -17th
[48] S. Wang, X. Zhang, Y. Zhang, L. Wang, J. YANG, and W. Wang, “A survey on International Conference on Smart Technologies. IEEE, Jul 2017.
mobile edge networks: Convergence of computing, caching and com-
[67] E. K. Markakis, K. Karras, A. Sideris, G. Alexiou, and E. Pallis, “Com-
munications,” IEEE Access, vol. 5, pp. 6757–6779, 2017.
puting, caching, and communication at the edge: The cornerstone for
[49] S. Parkvall, E. Dahlman, A. Furuskar, and M. Frenne, “NR: The new 5G building a versatile 5G ecosystem,” IEEE Communications Magazine,
radio access technology,” IEEE Communications Standards Magazine, vol. 55, no. 11, pp. 152–157, Nov 2017.
vol. 1, no. 4, pp. 24–30, Dec 2017.
[68] K. Zhang, Y. Mao, S. Leng, Q. Zhao, L. Li, X. Peng, L. Pan, S. Maharjan,
[50] Y. He, F. R. Yu, N. Zhao, and H. Yin, “Secure social networks in 5G and Y. Zhang, “Energy-efficient offloading for mobile edge computing in
systems with mobile edge computing, caching, and device-to-device 5G heterogeneous networks,” IEEE Access, vol. 4, pp. 5896–5907, 2016.
communications,” IEEE Wireless Communications, vol. 25, no. 3, pp.
103–109, Jun 2018. [69] Y. Ai, M. Peng, and K. Zhang, “Edge computing technologies for internet
of things: a primer,” Digital Communications and Networks, vol. 4, no. 2,
[51] H. Wang, J. Wang, G. Ding, L. Wang, T. A. Tsiftsis, and P. K. Sharma, pp. 77–86, Apr 2018.
“Resource allocation for energy harvesting-powered D2D communica-
tion underlaying UAV-assisted networks,” IEEE Transactions on Green [70] E. Bastug, M. Bennis, and M. Debbah, “Living on the edge: The role
Communications and Networking, vol. 2, no. 1, pp. 14–24, Mar 2018. of proactive caching in 5G wireless networks,” IEEE Communications
Magazine, vol. 52, no. 8, pp. 82–89, Aug 2014.
[52] X. Chen, L. Pu, L. Gao, W. Wu, and D. Wu, “Exploiting massive d2d col-
laboration for energy-efficient mobile edge computing,” IEEE Wireless [71] S.-C. Huang, Y.-C. Luo, B.-L. Chen, Y.-C. Chung, and J. Chou,
Communications, vol. 24, no. 4, pp. 64–71, Aug 2017. “Application-aware traffic redirection: A mobile edge computing imple-
mentation toward future 5G networks,” in 2017 IEEE 7th International
[53] Y He, J Ren, G Yu, Y Cai, "D2D Communications Meet Mobile Edge
Symposium on Cloud and Service Computing (SC2). IEEE, Nov 2017.
Computing for Enhanced Computation Capacity in Cellular Networks",
IEEE Transactions on Wireless Communications, vol. 18, no. 3, pp. 1750- [72] H.-C. Hsieh, J.-L. Chen, and A. Benslimane, “5G virtualized multi-access
1763. Feb 2019 edge computing platform for IoT applications,” Journal of Network and
[54] J. Zhang, W. Xia, F. Yan, and L. Shen, “Joint computation offloading Computer Applications, vol. 115, pp. 94–102, Aug 2018.
and resource allocation optimization in heterogeneous networks with
[73] B. P. Rimal, D. P. Van, and M. Maier, “Mobile edge computing empowered
mobile edge computing,” IEEE Access, vol. 6, pp. 19 324–19 337, 2018.
fiber-wireless access networks in the 5G era,” IEEE Communications
[55] N. Bessis and C. Dobre, Big data and internet of things: a roadmap for Magazine, vol. 55, no. 2, pp. 192–200, Feb 2017.
smart environments. Springer, 2014, vol. 546.
[74] J. Liu, H. Guo, H. Nishiyama, H. Ujikawa, K. Suzuki, and N. Kato, “New
[56] J. A. Silva, R. Monteiro, H. Paulino, and J. M. Lourenco, “Ephemeral perspectives on future smart FiWi networks: Scalability, reliability, and
data storage for networks of hand-held devices,” in 2016 IEEE Trust- energy efficiency,” IEEE Communications Surveys & Tutorials, vol. 18,
com/BigDataSE/ISPA. IEEE, Aug 2016. no. 2, pp. 1045–1072, 2016.

VOLUME 4, 2016 xv

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2938534, IEEE Access

[75] X. Huang, R. Yu, J. Kang, Y. He, and Y. Zhang, “Exploring mobile edge KOK-LIM ALVIN YAU (M’08–SM’18) received
computing for 5G-enabled software defined vehicular networks,” IEEE the B.Eng. degree (Hons.) in electrical and elec-
Wireless Communications, vol. 24, no. 6, pp. 55–63, Dec 2017. tronics engineering from Universiti Teknologi
PETRONAS, Malaysia, in 2005, the M.Sc. de-
[76] R. Solozabal, A. Sanchoyerto, E. Atxutegi, B. Blanco, J. O. Fajardo, and
F. Liberal, “Exploitation of mobile edge computing in 5G distributed
gree in electrical engineering from the National
mission-critical push-to-talk service deployment,” IEEE Access, vol. 6, University of Singapore in 2007, and the Ph.D.
pp. 37 665–37 675, 2018. degree in network engineering from the Vic-
toria University of Wellington, New Zealand, in
[77] S. Singh, Y.-C. Chiu, Y.-H. Tsai, and J.-S. Yang, “Mobile edge fog comput- 2010. Dr. Yau is currently an Associate Professor
ing in 5G era: Architecture and implementation,” in 2016 International with the Department of Computing and Infor-
Computer Symposium (ICS). IEEE, Dec 2016. mation Systems, Sunway University. He was the recipient of the 2007
[78] T. X. Tran, A. Hajisami, P. Pandey, and D. Pompili, “Collaborative mobile Professional Engineer Board of Singapore Gold Medal for being the
edge computing in 5G networks: New paradigms, scenarios, and chal- best graduate of the M.Sc. degree in 2006/2007. He is a Researcher,
lenges,” IEEE Communications Magazine, vol. 55, no. 4, pp. 54–61, Apr a Lecturer, and a Consultant in cognitive radio, wireless networks,
2017. applied artificial intelligence, applied deep learning, and reinforcement
learning. He serves as an Editor of the KSII Transactions on Internet
[79] I. A. Ridhawi, M. Aloqaily, Y. Kotb, Y. A. Ridhawi, and Y. Jararweh, “A col-
and Information Systems, an Associate Editor of the IEEE Access, a
laborative mobile edge computing and user solution for service compo-
Guest Editor of the Special Issues of the IEEE Access, IET Networks,
sition in 5G systems,” Transactions on Emerging Telecommunications
IEEE Computational Intelligence Magazine, Springer Journal of Ambient
Technologies, p. e3446, Jun 2018.
Intelligence and Humanized Computing, and a regular Reviewer for over
[80] M. Breternitz, K. A. Lowery, P. Kaminski, and A. Chernoff, “System and 20 journals, including the IEEE journals and magazines, the Ad Hoc
method for allocating a cluster of nodes for a cloud computing sys- Networks, the IET Communications, and others. He serves as a TPC
tem based on hardware characteristics,” Feb. 13 2014, US Patent App. Member and a Reviewer for major international conferences, including
13/568,395. ICC, VTC, LCN, Globecom, and AINA. He also served as the Vice General
[81] K. Kanai, K. Imagane, and J. Katto, “[invited paper] overview of multi-
Co-Chair of ICOIN’18, the Co-Chair of IET ICFCNA’14, and the Co-Chair
media mobile edge computing,” ITE Transactions on Media Technology (Organizing Committee) of IET ICWCA’12.
and Applications, vol. 6, no. 1, pp. 46–52, 2018.

[82] Z. Guan, Y. Zhang, L. Wu, J. Wu, J. Li, Y. Ma, and J. Hu, “APPA: An
anonymous and privacy preserving data aggregation scheme for fog-
enhanced IoT,” Journal of Network and Computer Applications, vol. 125,
pp. 82–92, Jan 2019.

[83] K. Jain and S. Mohapatra, “Taxonomy of edge computing: Challenges,

opportunities, and data reduction methods,” in Edge Computing.
Springer International Publishing, Nov 2018, pp. 51–69.

[84] E. Ahmed and M. H. Rehmani, “Mobile edge computing: Opportuni-

ties, solutions, and challenges,” Future Generation Computer Systems,
vol. 70, pp. 59–63, May 2017.

[85] K.Kaur, S.Garg, G.Kaddoum, M.Guizani, D.Jayakody, "A Lightweight and

Privacy-Preserving Authentication Protocol for Mobile Edge Comput-
ing", arXiv preprint arXiv:1907.08896, July 2019.
Celimuge Wu received his ME degree from
the Beijing Institute of Technology, China in
2006, and his PhD degree from The Univer-
sity of Electro-Communications, Japan in 2010,
where he is currently an associate professor.
His current research interests include vehicular
networks, sensor networks, intelligent transport
systems, IoT and edge computing. He is/has
been a TPC Co-Chair of Wireless Days 2019,
ICT-DM 2019, ICT-DM 2018, a track Co-Chair
of many international conferences including IEEE VTC 2020-Spring,
Najmul Hassan received an M.S. in computer
ICCCN 2019, and IEEE PIMRC 2016. He serves as an associate editor
sciences from Mohammad Ali Jinnah Univer-
of IEEE Access, IEICE Transactions on Communications, International
sity, Islamabad, Pakistan in 2010. He is currently
Journal of Distributed Sensor Networks, and MDPI Sensors.
a PhD scholar Department of Computing and
Information Systems, Sunway University. Prior
to starting his doctorate, he was working as
visiting faculty in Department of Computer Sci-
ence, COMSATS University Islamabad, Attock
and Abbottabad Campus. His research areas are
Internet of Things, edge/cloud communication,
wireless body area networks, airborne internet and wireless sensor
networks.

xvi VOLUME 4, 2016

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.