Carvalho 2021

Computing (2021) 103:993–1023
https://doi.org/10.1007/s00607-020-00896-5
SURVEY ARTICLE
Edge computing: current trends, research challenges and

future directions
Gonçalo Carvalho1 · Bruno Cabral1 · Vasco Pereira1 · Jorge Bernardino1,2
Received: 28 July 2020 / Accepted: 22 December 2020 / Published online: 18 January 2021
© The Author(s), under exclusive licence to Springer-Verlag GmbH, AT part of Springer Nature 2021
Abstract
The edge computing (EC) paradigm brings computation and storage to the edge of
the network where data is both consumed and produced. This variation is necessary to
cope with the increasing amount of network-connected devices and data transmitted,
that the launch of the new 5G networks will expand. The aim is to avoid the high
latency and traffic bottlenecks associated with the use of Cloud Computing in net-
works where several devices both access and generate high volumes of data. EC also
improves network support for mobility, security, and privacy. This paper provides a
discussion around EC and summarized the definition and fundamental properties of the
EC architectures proposed in the literature (Multi-access Edge Computing, Fog Com-
puting, Cloudlet Computing, and Mobile Cloud Computing). Subsequently, this paper
examines significant use cases for each EC architecture and debates some promising
future research directions.
Keywords Edge computing · Fog computing · Cloudlet computing · Multi-access

edge computing · Mobile cloud computing
Mathematics Subject Classification 68-02
B Gonçalo Carvalho
gcarvalho@dei.uc.pt
Bruno Cabral
bcabral@dei.uc.pt
Vasco Pereira
vasco@dei.uc.pt
Jorge Bernardino
jorge@isec.pt
1 Department of Informatics Engineering, CISUC, University of Coimbra, Coimbra, Portugal

2 Polytechnic of Coimbra, ISEC, Coimbra, Portugal
123
994 G. Carvalho et al.
1 Introduction
As the number of network-connected devices continues to increase at a fast pace,

the shortcomings of traditional network architectures, where data is sent and received
to/from the core of the network, become visible. These devices, supported by increas-
ingly faster access networks, not only demand lower delay and higher speed when
accessing data from the network but also generate an increasing amount of data to
send. Also, because of the hardware restrictions of some edge devices, especially those
mobile, there is a high demand for offloading tasks, creating additional bottlenecks in
an already congested network. The proliferation of Internet of Things (IoT) devices,
and the foreseeable bandwidth demand caused by the new 5G networks devices and
applications, demand alternative solutions and architectures to deal with these new
requirements.
To respond to the bottlenecks of traditional Cloud Computing (CC) networks, Edge
Computing (EC) architectures emerged. The goal was to optimize the network and its
applications by bringing data production closer to data usage, minimizing the need
for long-distance communications between edge clients and servers. The result is a
reduction in latency and bandwidth usage, a general improvement in connectivity, a
more efficient network operation and service delivery, better scalability, faster access
to external computation and storage. Also, it is easier to achieve higher security, data
privacy, and reliability can. Generally, EC can be defined as an umbrella term gathering
all architectures that aim to bring computation and storage resources closer to the end-
users and devices [93].
In recent years, several authors published surveys concerning EC architectures, such
as [1,2,19,19,30,37,55,65,67,69,74,77,81,85,87,89,117], and [125]. Because these are
valuable research papers, we target them as resources for more detailed information
regarding each EC architecture. Also, this paper focus on EC use cases and future
research directions. This paper makes the following contributions:
– Present a clear and deep analysis on the main features of the EC paradigm;
– Provide the definition and major properties of the essential EC architectures.
– Present relevant use cases for each architecture to display the possibilities of EC;
– Compare different architectures by analysing each in generic Smart City scenarios;
– Discuss existing open issues and future research directions.
The rest of this paper is organized as follows. Section 2 provides a thorough review
of the Edge Computing paradigm. In Sect. 3, the definitions and essential properties of
the main EC architectures proposed in the literature (Multi-access Edge Computing,
Fog Computing, Cloudlet Computing and Mobile Cloud Computing), are discussed.
Section 4 presents use cases for each architecture, also identifying how to deploy these
in Smart Home, Building, and City use cases. Section 5 discusses future research
directions and open issues concerning the paradigm and its architectures. Finally,
Sect. 6 presents the main conclusions and future work.
123
Edge computing: current trends, research challenges and... 995
2 The edge computing paradigm
Previous to Edge Computing (EC), Cloud Computing (CC) arrived with many new
capabilities, such as on-demand services and applications, scalability, efficient data
storage, and management [51], fault-tolerance, easy management, and stable pric-
ing policies [77]. However, the evolution of technology also led to a fast rise in the
amount of data generated, processed, and stored. The result was that CC solutions
exhibited problems, such as undesirable latency, less mobility support, the inability of
context awareness, data transmission overhead, loss of privacy, and limitations on the
connectivity between Cloud and end devices [69,77].
A new computing paradigm emerged to tackle these challenges, known as Edge
Computing. EC was a term first coined by Akamai in the late 1990s, when he intro-
duced content delivery networks (CDNs) to speed up web performance [69]. Without
trying to replace CC, but acting as its complement,EC solutions support CC archi-
tectures by serving geographical and latency-sensitive applications. CC also supports
EC with extensive storage and computing capacities [82]. EC enables local distributed
data processing by taking advantage of nearby devices, going beyond simple computa-
tion offloading to the Cloud [69]. EC intends to provide a better quality of service and
experience to users by efficiently trimming response time and throughput, enabling
real-time applications and services. EC also aims to afford high scalability, reliability,
and fault-tolerance while increasing privacy and security [40]. Additionally, EC sys-
tems intend to drive computation proficiency to the edge of the network, eliminating
the previous problems of a centralized architecture [34].
According to Khan et al. [65] there are several requirements to enable EC:
– Billing mechanisms which must assess the resource availability, the frequency of
resource usage, and the duration of resource usage;
– Joint business model for management and deployment aiming at high performance,
communal management and deployment, and low cost services;
– Real time application support with reliable edge node selection, dedicated com-
puting resources, and low latency;
– Redundancy and fail-over capabilities regarding high availability of services,
resilience and reliability;
– Resource management analyzing the processing power, network resources, and
power consumption;
– Scalable architecture through resource virtualization, blockchain based trust estab-
lishment, and automated Edge-IoT orchestration;
– Security by authentication processes, confidentiality, and data integrity.
Software solutions for virtualization, such as Virtual Machine (VM), containers, and
their management or migration tools, have been active enablers of EC systems. More
recently, software-managed networks also allow further exploration of the potential
of EC by facilitating technologies, such as Software-Defined Networking (SDN),
Network Function Virtualization (NFV), and Overlay network [54]. SDN detaches
the control plane which establishes where the traffic is sent, from the data plane which
forwards the traffic to the destination determined by the control plane. NFV disjoints
networking functions, such as routing and fire-walling, from the hardware, and allows
123
each of the functions to run on a VM. An overlay network is a virtual network within
the physical network providing additional services.
EC works as a distributed computing standard, where collected data is processed
closer to users, thus avoiding uploading a significant amount of data to the cloud center
[111]. Nevertheless, there is no consensus on the limits of each EC architecture pro-
posed in the literature, and many authors use the terms interchangeably. For example,
Varshney and Simmhan [114], consider that Fog Computing and Cloudlet Computing
are synonyms. The European Telecommunications Standards Institute (ETSI) [86],
named Multi-access Edge Computing as EC’s standard architecture but uses both
terms to address the concept of a Cloud-based Information Technology service envi-
ronment. Moreover, Yi et al. [123], stated that Fog Computing is a combination of
Mobile Cloud Computing and Multi-access Edge Computing, directed to the context
of Internet of Things (IoT). Also, Chiang and Zhang [24] stated that despite fog and
edge are used interchangeably, the term fog is wider than the later. Hu et al. [55] and
Chiang et al. [23], both state that Fog and Edge do not belong under the same umbrella
of EC.
Additionally, the term architecture, used by Chiang and Zhang [24] and Hu et al.
[55], is not used globally by the research community. Wang et al. [115], specified that
Mobile Edge Computing, Fog Computing, and Cloudlet Computing are schemes of
EC. Dinh et al. [30], referred to Mobile Cloud Computing as an infrastructure. Mahmud
et al. [74], Ren et al. [89], and Yousefpour et al.[125], used the term paradigm. Some
authors, such as [46,69], and [34], to name a few, considered EC as a single architecture,
to provide the same global benefits associated with the EC paradigm.
Notwithstanding all different points of view, EC is referred as all computing solu-
tions, architectures, schemes, and so on, that are not within the core of the network.
Thus, even with the different opinions around the limits and layers of Edge or Fog
Computing, in this paper, it is addressed as EC systems. Figure 1 portrays the Edge-to-
Cloud continuum and deployment scenarios for each architecture (Multi-access Edge
Computing, Fog Computing, Cloudlet Computing, and Mobile Cloud Computing).
From the top, centralized Cloud services compose the core of the network, providing
storage and, sometimes, computation to the end users through a Wide Area Network
(WAN) connection. Away from the core of the network, in an area close to the final
user or application, is the domain of what is called the Edge of the network.EC is the
global paradigm that encompasses all architectures within edge scope, and it registers
distributed computation, storage, and networking outside of the centralized Cloud
solutions.
At the edge level, the four fundamental architectures described in the literature
are Multi-access Edge Computing, Fog Computing, Cloudlet Computing, and Mobile
Cloud Computing. While their boundaries are not strict and overlapping concepts
exist, there are enough characteristics, which are addressed next, to enable an ade-
quate differentiation between them. Nevertheless, other architectures were found in
the literature, such as Mobile Grid Computing [49], Edge-Centric Computing [42],
Dew Computing [118], Mobile Crowd Computing [38], Mist Computing [124], and
Transparent Computing [130]. However, we perceive them as extensions or derivations
to the previously mentioned architectures.
123
Fig. 1 Edge computing deployment scenarios
Between end users, or devices, and the Cloud, the EC paradigm does not intend to
replace Cloud services, but to extend them, providing powerful computing resources
near the end users, with lower network traffic and lower latency, while providing a
better user experience. Nevertheless, there is always the possibility of an edge node to
access the network core if needed. Hence, the arrows to the fog nodes and the cloudlet
depicted in Fig. 1 may return directly to edge devices or continue and connect to the
core of the network. The dashed arrows in grey, between the fog node and the cloudlet,
and between the mobile edge devices and the cloudlet, are not mandatory to the
architectures, i.e. each architecture may exist independently. However, the connection
between different edge architectures is possible and can contribute to enhancing the
system. As an example, fog nodes may connect to cloudlets to improve the computing
and storage resources, without the need to connect to the more distant Cloud resources.
End users or devices complete the architectures and are placed at the lower level of
the network, further from the core. Some devices that belong to this layer are smart-
phones, cameras, intelligent vehicles, smart wearables, IoT sensors, and actuators.
These devices access/produce data and may also communicate among them using
short-range communications.
When compared to CC, EC presents many distinct characteristics. EC applications
are normally user-driven and use distributed architectures, in contrast with the data-
driven applications and centralized architectures of CC. Benefiting from its location
closer to the edge of the network where data is produced and consumed, EC has
reduced bandwidth consumption and latency, which also promotes energy savings
and both quality of service and experience improvements. In EC, the resources are
few, widespread, and with high heterogeneity, contrasting with the many resources
provided by CC, which are locally clustered and similar. Regarding scalability, EC
architectures are very scalable in the number of devices that are supported, meaning
123
Table 1 Computing characteristics comparison between cloud computing (Cloud) and edge computing
(Edge) paradigms
Characteristics Cloud Edge
Applications Data-driven User-driven

Architecture Centralized Distributed
Availability High Average/high
Bandwidth utilization High Low
Computing resources High Moderate
Distance from users Far Close
Energy consumption High Low
Geo-distribution of computing resources Locally clustered Widespread
Heterogeneity of computing hardware Low High
Latency High Low
Number of computing resource locations Many Few
Quality of service and experience Medium High
Resources at individual locations Many Few
Response time High Low
Scalability Medium High
Service access Core Edge
Service type Global Local
Storage capacity High Low
Virtualization Heavy Light
that the architecture must always be prepared to register an increase in the number of
connected devices, however, adding complexity to the devices’ management process.
Another limitation of these edge architectures is the storage capacity, and that is why
CC continues essentially for long-term data storage. Since CC is based on centralized
data centers, these have high availability and abundant computing resources. Avail-
ability in Edge solutions range from high, in Multi-access Edge Computing, to average
in the remaining architectures, and is moderate concerning the computing resources.
Regarding the service access, in CC it is made through the core of the network, while
EC may hold different possibilities, such as through the edge of the Internet in Multi-
access Edge Computing, by connected devices at the edge in Fog Computing, through
the core in Mobile Cloud Computing, and by resource-rich computers at the edge of
the network in Cloudlet Computing. Finally, CC allows heavyweight virtualization,
while EC virtualization has some restrictions because of lower processing capabilities.
Table 1 summarizes the comparison between CC and EC, regarding their most relevant
characteristics.
123
3 Edge computing architectures
This section provides an overall analysis of each Edge Computing (EC) architecture
shown in Fig. 1, through their definitions and particular properties. Along the path
between data sources and cloud data centers, EC devices are considered as any com-
puting and network resource [96]. Each architecture may behave autonomously or with
a Cloud structure. In the first case, data is processed locally at the edge of the network,
only accessing the Cloud when strictly necessary, minimizing network overhead, and
improving security and privacy [34].
Multi-access Edge Computing: formerly known as Mobile Edge Computing,

refers to computing at the edge of a network and the European Telecommunication
Standards Institute (ETSI) [86] defined its five major characteristics: on-premises,
meaning that Multi-access Edge Computing platforms can run isolated from the rest
of the network; proximity; lower latency; location awareness, because edge-distributed
devices use low-level signaling, for information sharing; and network context infor-
mation. These features extend the initially proposed EC architecture, cited by some
authors as ([46,69], and [34]), adding the possibility of wireless access and differen-
tiated virtualization technologies. The Reference Architecture for Multi-access Edge
Computing is detailed in [35].
Fog Computing: Vaquero and Merino [113], defined Fog Computing as a ”scenario
where a huge number of heterogeneous (wireless and sometimes autonomous) ubiq-
uitous and decentralized devices communicate and potentially cooperate among them
and with the network to perform storage and processing tasks without the intervention
of third-parties”. Fog Computing arose to cope with the vast amount of Internet of
Things (IoT) devices and big data, as an extension of Cloud Computing (CC) and
designed for real-time low-latency applications. The boundaries of Fog Computing
are hard to establish, and the possibilities of implementation are several, ranging from
IoT to cloud services usage. Also, the Reference Architecture for Fog Computing is
portrayed in [25].
Cloudlet Computing: the cloudlet concept was first proposed by Satyanarayanan

et al. in 2009 [94]. The author stated it as a “trusted, resource-rich computer or cluster
of computers that are well-connected to the Internet and available for use by nearby
mobile devices, offering computing and storage resources”. Cloudlets aim to endorse
flexibility, mobility, scalability, and elasticity. Cloudlets are discoverable, generic,
stateless servers located in single-hop proximity of mobile devices. Thus, benefiting
from Virtual Machine (VM), and can operate disconnectedly from Cloud services
[26]. Acting as a “data center in a box” [115], and through a high-bandwidth network,
cloudlets resemble a set of virtualized high-performance computers [69], with lower
latency, to geographical close edge devices. The Reference Architecture for Cloudlet
Computing is presented in [16].
Mobile Cloud Computing: mobile devices are increasingly becoming an essen-

tial part of human life as practical and convenient communication tools that are not
123
Table 2 Properties of edge computing architectures
Multi-access edge Fog computing Cloudlet Mobile cloud

computing computing
Ownership Telecommunication Private entities or individuals

companies
Deployment Network edge, Edge to near-edge, Network edge Network edge,
Edge devices devices devices
Hardware Routers, switches, gateways, servers, access points Servers, user
devices
Services Virtualization Virtualization,
provided others
Network N-tier, decentralized, distributed
architecture
User mobility Yes
management
Latency Low High
Connection to Yes or No Yes
the Cloud
Location Yes or No Yes Yes or No
awareness
Availability High
Scalability High Higher
Target user IoT/mobile Mobile
Distance to users Low High
Working Outdoor/indoor Outdoor
environment
Jitter Low High
Computational High Low High Low (mobile),
power high (Cloud)
Storage capacity High Low High Low (mobile),
high (Cloud)
bounded by time and place. According to the Mobile Cloud Computing Forum, ref-
erenced by Yi et al. [123],“Mobile Cloud Computing at its simplest, refers to an
infrastructure where both the data storage and data processing happen outside of the
mobile device. Mobile Cloud applications move the computing power and data storage
away from mobile phones and into the Cloud, bringing applications and Mobile Com-
puting to not just smartphone users but a much broader range of mobile subscribers”.
Moreover, Mobile Cloud Computing’s Reference Architecture is illustrated in [30].
To contribute to a better understanding of the different EC architectures, Table 2
summarizes the fundamental properties of each architecture. To clarify, when Mobile
Cloud Computing uses a cloudlet in its architecture, it suits Cloudlet Computing, hence
the properties are the same as the ones presented in the respective column.
123
Table 2 continued
Multi-access edge Fog computing Cloudlet Mobile cloud

computing computing
Power Low Low Low Low (mobile),

consumption high (Cloud)
Hardware Wi-Fi, 3G, 4G, Wi-Fi, LAN Wi-Fi, LAN, Bluetooth,
connectivity LTE, LAN, WAN Wi-Fi,3G, 4G,
WAN TE
(mobile),LAN,
WAN (Cloud)
Power source Battery/direct power/renewable energy Battery (mobile),
direct power
(Cloud)
Node mobility Low High Low High (mobile),
low (Cloud)
Standardization ETSI OpenFog OpenEdge NIST
consortium,
IEEE
Optimization Processing capac- Latency Processing
problem ity/storage/latency capacity/storage
Rentable Yes or no Yes or no Yes No
Service access Through servers at Through Through Through data
the edge of the connected resource-rich centers at the
internet devices at the computers at the core of the
edge edge of the internet
internet
Server location Base stations Can be installed at Near mobile Installed in large
the edge or in devices at coffee dedicated
dedicated shops, public buildings
locations transportation
stations
4 Edge computing use cases
In this section are provided some use cases found in the literature for each Edge
Computing (EC). architecture introduced in previous sections. To provide an easy
comparison of the strengths and weaknesses of the different architectures, each will
apply to a Smart City scenario (including sub-scenarios such as Smart Home or Smart
Building) because this scenario has many possibilities for EC architectures. Each
architecture provides different implementations under this use case. Table 3 displays
relevant use cases for each EC architecture.
Multi-access Edge Computing The scenarios that take advantage of Multi-access

Edge Computing benefit from the use of multiple access technologies to access the
network and locally gathered data. The static or mobile edge devices in this architecture
have access to stand-alone servers at base stations, which may also collect data from
other surrounding sources. By reducing or even avoiding access to data on the Cloud,
123
Table 3 Edge computing use cases for multi-access edge computing (MEC), fog computing (Fog), Cloudlet, and mobile cloud computing (MCC)
1002
Use case MEC Fog Cloudlet MCC
123
Augmented reality [10,77,95,119,125] [19,123] [36,55] [36] [31,117]
Client-controlled Cloud storage/services [24] [117]
Computation offloading [28,72,90,91, [71] [62,107,108, [9,29,41,48,
96,108,110] 131] 52,76,77]
Content aggregation/delivery [10,28,58,108] [123] [100]
Crowd sensing/sourcing [116] [24] [98,121] [117,120,125]
Healthcare [1,58,80,84] [33,51,55,59, [20,78,109] [87,117,125]
125] [30,75,109]
Mobile big data analytics [1,10,65,108] [65,123] [65]
Mobile gaming [50,55] [30,117]
Mobile learning/commerce [30,87]
Quality of service and [72,103] [5] [21] [120]
quality of experience
improvement
Real-time analytics [65,72] [55,65,125] [65] [48]
Smart city [96,108] [25,55,59,61,125] [26,44,99,109] [18,102]
Smart grid [1,58] [19,105,125] [55] [92]
G. Carvalho et al.
Table 3 continued
Use case MEC Fog Cloudlet MCC
Smart home/ building [1,96] [25,50,55,59, [45,79,122] [128]

105,126]
Smart traffic light system connected vehicles [1,11,58,77, [15,50,125] [117]
84,91,95,125, [19,25,55,59,71]
129,132]
Video analytics/cameras network [1,72,77,91,96,108, [25,50,123,125] [12,77,108] [4]
110,125]
Wind farm/ocean monitoring [1] [14]
Wireless sensors and actuators networks [1] [15,105]
Edge computing: current trends, research challenges and...
123
1003
it enables real-time responses to local events and aids the overall management of a
system, which translates into a better overall user experience. These conditions are
especially compelling to telecommunication operators that, by using this architecture
(mostly with cellular networks), can provide better services to its users at a controlled
cost.
Typical scenarios found in the literature that benefit from the use of Multi-access
Edge Computing are Smart traffic light system/connected vehicles [11]. Content deliv-
ery and caching, video analytics, mobile big data analytics, smart grid [1]. Computation
offloading, data preprocessing, context-aware services [110]. Face/speech recognition
or image/video editing, real-time information [72]. Intelligent transportation systems
[11], and augmented reality [119], but only in situations where the number of foresee-
able users is limited not to overload the server.
Because of its characteristics, Multi-access Edge Computing can facilitate sev-
eral Smart City services, especially those that require low processing and immediate
access to local information. The authors in [1] discuss Smart Building Control where
sensors scattered in the building monitor and control its environment, regarding tem-
perature, gas level, or humidity, which allows reacting quickly to any anomaly. Another
Smart City use case is to perform video analytics in case of a missing person, crim-
inal monitoring, or hazard situation, which would prevent the network’s overload
with data backhaul by processing data locally, thus guaranteeing low latency [108].
Still, within in the Smart City environment, the authors of [95,129,132], addressed
connected vehicles (Fig. 2) which use images or video analysis to transfer or receive
information to other vehicles about traffic congestion or road conditions to achieve
route prediction and collision warning applications. These have the aid of Roadside
Units (RUs) to push Vehicles to Everything (V2X) applications, data, and services
from the central Cloud to the edge of the network, thus increasing efficiency [132].
Multi-access Edge Computing servers could be deployed in RUs, in base stations, or
could integrate efforts with the RUs, bringing data and analytics applications closer
to the vehicles, enabling applications acceleration over the vehicles [129]. The near
real-time communication improves road safety through diminishing driver’ reaction
time, hence avoiding accidents [95]. The emerging self-driving and autonomous cars
will benefit from Multi-access Edge Computing by the provided assisting information
[108].
Fog Computing Fog Computing has less resourceful devices than the previous
architecture. However, the inter-connectivity and collaboration of these devices allow
running powerful applications through a more compelling collection of resources. Fog
Computing also benefits from the nodes’ distribution across the environment, each pro-
viding local information and distributed computational and storage resources. Being
an affordable architecture, much more than Multi-access Edge Computing servers or
cloudlets, for everyone to deploy, it has almost limitless applications.
The following are some more significant use cases in Fog Computing found in
literature: healthcare services [51], wireless sensors, and actuators networks [105]
and surveillance [22]. The following authors [45,79,122] claim that edge devices
resources cannot deal with the amount of produced data, computation, and storage
needs for Smart Home, Building, City use cases. Thus, present Fog Computing as
123
Fig. 2 Connected vehicles for a smart traffic system using multi-access edge computing
the solution for the mentioned limitations, providing a home security application, as an
example, where the devices would work as clients, and the servers would be deployed
in Virtual Machine (VM) at fog nodes, allowing real-time analytics.
In the wider scope of a Smart City, the authors in [45] discuss noise pollution
mapping for better city planning and diminish health problems connected to this
pollution, monitoring urban drainage networks to prevent urban floods in extreme
weather occurrences, and smart streets where the fog devices would provide real-time
street status monitoring and also services to the citizens. Yu et al. [126], developed
an energy prediction and management system in smart homes and buildings. Jia et al.
[61], optimized street lighting through a fog-based smart street lamp, which imple-
ments dynamic brightness adaptation, autonomous management, and self-governing
alarm on irregular states. Javadzadeh and Rahmani [59] survey Fog Computing appli-
cations in Smart Cities, categorizing their analysis in municipal services, such as waste
and water management, parking and transportation, and lighting; smart citizen; smart
education; smart healthcare; smart building; smart energy; and smart governance.
Using Fog Computing for Smart Cities, Fig. 3a displays a method applicable in a
surveillance scenario [22]. In this example, the authors used a three-layer surveillance
architecture. Collecting data through smart devices and sending it to the end users in
the user or surveillance application layer. Then the Fog Computing layer processes
and the video and does the real-time data processing and storage, enabled by devices
such as tablets, smartphones, laptops in a police car, or onboard computing devices on
a drone. The system preprocesses videos to remove irrelevant data, increase privacy,
and also to diminish latency and data traffic in a future upload to the Cloud. It may
require the Cloud Computing layer for complex tasks such as data analytics.
Cloudlet Computing The use cases that need higher computation performance
than the provided by fog nodes but also need lower latency than Multi-access Edge
123
Fig. 3 a A smart surveillance architecture using fog computing; b a crowd-sourcing infrastructure using
cloudlet computing; c Augmented reality application using mobile cloud computing
Computing should contemplate the application of Cloudlet Computing for a higher

quality of experience and service. Cloudlets additionally benefit scenarios that require
access to data centers and can build clusters of cloudlets to achieve high-performance
computation and storage.
According to the literature, Cloudlets benefit use cases such as video stream, image
processing, file editing [108], face recognition, computation offloading [77], inter-
active media, speech recognition, natural language processing [11], and augmented
reality applications. In the latter, cloudlets aid the user’s real-world perception by
including virtual elements into the scenario related to one’s context and movement
[43]. Using a cloudlet to support an augmented reality application is limited to sce-
narios where the required processing capabilities and the number of users are fit with
the edge resources.
In the Smart Home, Building, City use case, offloading, and processing data in
cloudlets [109]. These devices can communicate with each other and perform coor-
dinated functions to serve the city-systems requirements. According to Gedeon et al.
[44], available studies have neglected the cloudlet’s location. So the authors proposed
to deploy cloudlets in selected infrastructures across the city environment, such as
cellular base stations, routers, and street lamps, thus resulting in low to none visual
pollution and providing coverage for mobile users to access these resourceful devices.
Yet another solution for Smart City concerns public transportation commuters by
deploying a cloudlet inside the bus allowing for augmented reality or computation
offloading within the trip with no disconnections and ensuring high quality of ser-
vice [11]. In a Smart City crowd-sensing scenario, Simoens et al. [98] introduced
GigaSight (Fig. 3b), which is a hybrid cloud architecture for scalable crowd-sourcing
of video from mobile devices. GigaSight labels the stored videos from the mobile
device in a local cloudlet for adequate uploading, downloading, and processing while
preserving privacy concerns.
123
GigaSight proposes the effortless upload of video from end users, thus enabling
efficient and quick search and processing of similar videos, since users uploaded these
from nearby locations, and it does all the processing locally. The user can query
GigaSight using a SQL type instruction, being able to find thieves, missing children,
or pets, for example.
Mobile Cloud Computing Mobile Cloud Computing applies for cases where sev-
eral users with different geographic locations need to have the same information.
Cloud data centers are the best solution because these efficiently host many clients at
the same time. In mobile gaming, for example, it connects many users within the same
game and from any place. Thus Cloud resources and solutions are necessary to share
all information and the game engine simultaneously by everyone while also allowing
players to offload the game processing.
Some use cases of Mobile Cloud Computing found in the literature are mobile gam-
ing, mobile commerce, mobile learning [30], mobile social media [87], multimedia
sharing, and location-based services [117]. Mobility, disaster recovery, waste manage-
ment, tourism, and healthcare by combining the user physical activity with medical
archives would enable smart and connected medical devices to present information
for medical diagnosis [75].
As for the use case of Smart Home, Building, City using crowdsensing as social
media sharing, providing data for analysis to assess a crucial point in the occurrence of
a disaster [117]. Also, traffic data sensing providing important information for trans-
portation managers by text mining social media posts. Also, using the end users’ smart
devices location services and providing feedback on waiting times, or information to
the transport companies to spread more vehicles or redirect others to the more crowded
locations [18]. Moreover, crowdsensing social media networks would enable not only
monitoring but also manage accident situations. Dolezal et al. [31], used an Aug-
mented Reality application (Percipio, Fig. 3c) in Prague, Czech Republic, to assess
the benefits of offloading computation. The information displayed helps tourists to
know what they are looking at and to discover other attractions to see next. Even if
the use of a cloudlet would decrease latency, Cloud data centers are more prepared to
cope with many users connecting to the same application, thus being an overall better
approach.
Table 4 presents the primary strengths and weaknesses of the different architectures
for the Smart Home, Building, City use case. Comparing the architectures, the cost
of deployment is higher in Multi-access Edge Computing because of the necessity of
dedicated edge servers at base stations. Solutions using cloudlets are also expensive
since these imply resource-rich edge nodes built to serve a few geographically close
clients [115].
Regarding raw processing capabilities, the most powerful are the cloudlets and, on a
lower scale, Multi-access Edge Computing, as the number of servers is more restricted.
In Fog Computing, the processing power of the network depends on the number and
individual capacities of the devices at a specific time and area. Finally, Mobile Cloud
Computing is as powerful as the Cloud it connects to, but its performance heavily
depends on the network conditions to access it.
123
Table 4 Strengths and weaknesses of the different architectures for the smart home, building, city use case
Strengths Weaknesses
Multi-access edge computing Narrow to wide scope Expensive servers

of implementations location limitations
powerful resources
one hop away
Fog computing Interconnected Not so powerful
devices may result devices, and more
in resourceful narrow scope of
processing implementations
capabilities cheap compared to
and easy multi-access edge
deployment computing or
Cloudlets
Cloudlet computing Narrow to wide scope Expensive
of implementations
powerful resources
easy deployment in
terms of locations
Mobile cloud computing Cheap deployment Limited solutions
highly mobile latency and device
solutions crowd battery concerns
processing narrow processing and
scope of solutions storage
shortcommings
Despite the several and distinguished possibilities of the provided use cases, each
deploys its independent resources. If possible, the use of the unique resources made
available by each architecture in a complementary way would be more efficient. In
a smart home environment, the resources of a fog node would provide real-time fea-
tures to the system but could connect to Cloud services for long-term data storage.
Using cloudlets in a neighborhood environment would allow for edge servers at base
stations (Multi-access Edge Computing) to perform powerful computations and real-
time responses to the every-day scenario situations, such as traffic management. Cloud
resources would allow performing data analytics, running machine learning algo-
rithms, and storing data. Mobile Cloud Computing would benefit from crowdsensing
data made available from mobile devices.
5 Future research directions
Despite the continued research and improvement, in both academic and industrial
systems, Edge Computing (EC) still faces crucial challenges. In this section, we explore
some gaps in the current state-of-the-art and provide critical research directions for
future work on edge technologies.
Table 5 displays a chronologic overview of the EC research topics, derived from
the literature review, which allows us to understand the evolution through time of
123
the distinct fields addressed by the research community. As some issues are recurrent
over the years, it was only considered each topic once in the year the original authors
denoted it. Each research topic includes a reference.
Naturally, some mentioned issues have already been tackled but many continue to
be open for future research, such as scalability, mobility, security and privacy, device
heterogeneity, computation offloading, reliability, resource management, context-
awareness, and caching. The ever-evolving computer associated technologies and its
fast development culminate in keeping some of them open as new challenges arise or
they require better solutions. It is especially true when considering the advancements
in network speed and the devices’ resources, i.e. processing and storage. The launch
of 5G networks with faster and concurrent connections, the improvement of network
devices that enable more complex in-network computation and storage will allow
us to process complicated algorithms with better performance and higher complex-
ity. 5G networks will be the building block for EC systems, particularly in real-time
environments that require on-time decisions, notwithstanding that it will forge new
challenges to edge environments. In the next sub-sections, are elected open issues that
may provide excellent opportunities for future research by the scientific community.
5.1 Computation offloading
If an application requires high computational processing, or if the device running

and application has not the required processing resources, then computation offload-
ing, from end devices to edge nodes or even amongst edge nodes, is the best solution.
The goal is to extend battery life and enhance performance [108], aggregating external
computing resources and storage capacity to its strengths [69]. Also aiming to improve
reliability, because data from mobile devices is backed-up into edge nodes [9], thus
upgrading mobile devices’ computing capacities [85]. Computation offloading and
dynamic task partitioning are decisive in the EC environment. These techniques allow
better management of the trade-offs between energy consumption and system perfor-
mance [63], in the decision of handling the task locally or by offloading it [3]. Thus
resource discovery is also a crucial part of the offloading decision. Retrieving the
load and characteristics of the edge nodes to decide which one can run a task more
efficiently, taking into account characteristics, such as energy spent, the access cost,
or throughput, is an example of the decisions made in the edge. There is a multiplica-
tion of edge devices and services across the network, raising scalability issues, which
inadequate utilization of the resources may exacerbate [117]. Dealing with multiple
access requests from end users to a specific utility will create problems for the service
provider [79], which may be intensified by rush-hours demands. Furthermore, prepar-
ing EC architectures for the increase or decrease of available resources and users [83],
without compromising the edge devices network quality of service.
Therefore, how, when, and where behind the decision to offload a task is critical.
Choosing the node/device to receive the offloading process is hard. Even more, if
the decision implies multi-criteria optimization, such as increased computation per-
formance, saving battery lifetime, and low response time. The multiple parameters
involved make it a problem where Artificial Intelligence may present benefits. When
123
Table 5 Chronologic evolution of edge computing research topics
Year EC research topics Papers
2011 Scalabihtv [57]

Heterogeneity
Mobility
Awareness and adaptability
Security, privacy and trust
2012 Software architecture [27]
Power consumption on real devices
Synchronization
Power constraints
2013 Low bandwidth [30]
Network access management
Quality ol service
Pricing
Standard interface
Service convergence
2014 Computation or task offloading [87]
Mobile sensing
Mobile big data
Mobile Cloud federation
Mobile advertisement
Service oriented computing
2015 Task-oriented mobile services [117]
Human-driven distributed systems [42]
Edge architectures and middleware
2016 Reliability [26]
Resource management [50]
Latency requirements [24]
Resource-constrained devices
Cyber-physical system
Uninterrupted services with intermittent
Connectivity to the Cloud
Incentive for devices participation
Convergence and consistency
Standard protocol [2]
Simulation platform
123
Table 5 continued
Naming [97]
Context-aware resource service provisioning [74]
Sustainable and reliable fog computing
Inter-operable architecture of fog nodes
Distributed application deployment
Multi-tenancy support
Tools for fog simulation
Programming languages
2017 Data abstraction [96]
Service management
Optimization metrics
Interface (confidentiality,integrity, availability) [1]
Energy management [55]
Efficient deployment [95]
User mobility and transparency
Availability
Fog-cloud inter-working
Protocols and standardization [8]
Networking
Enabling real-time analytics [115]
Distribution and management of multi-access edge computing resources [72]
Traffic paradigm imposed by coexistence of offloaded data and conventional data
Concept validation
Caching [47]
Deployment strategies and orchestration
Data dissemination
Service enhancements: quality of experience [108]
Fairness [110]
2018 Data integrity [127]
Failure management [79]
Application service management and application modeling
Virtualization of fog devices [81]
applied to the computation offloading decision, it should answer questions such as

Should the task be offload or not? Full offload or fragments? Which nodes to choose
for the offloading process? When to perform the offloading process? Will offload
improve performance, save time, and battery life?
Machine Learning techniques, such as neural networks by achieving higher accu-
racy in the offloading decision in these dynamic environments will have a crucial
contribution. Lowering data traffic, reducing latency, minimizing communication bot-
123
Table 5 continued
Hardware technologies [69]

Human activity awareness
Resiliency
2019 Execution models [39]
Economy
Intelligent things
Mobile fog computing [125]
Fog service level agreements
Platform as a service for fog computing
Mobile and green fog computing
Resilient fog system design
Compatibility [70]
Edge dateway [88]
Adge analytics
Digital twin
User’s trust on EC svstems [65]
Flexible networked operating system [89]
API standards and development
Virtualization for network computing
2020 Dependability in EC [6]
Edge artificial intelligence [82]
Distributed recommender systems [106]
tlenecks, assuring better reaction times, providing higher adaptability and portability,
thus reducing costs and error probability, are some among others. Despite the increas-
ing number of papers regarding this subject, there are still research opportunities in
computation offloading decision optimization.
5.2 Edge computing and 5G
The impact of 5G networks in EC is still an open issue. Although ultra-low latency and
higher bandwidth will help to enhance edge solutions, the full repercussion is not yet
clear. 5G networks will potentiate an explosion of the number of IoT devices, increas-
ing their heterogeneity and security issues while demanding more data processing
from the network and increased device cooperation. Also, credentials management in
IoT environments is imperative, considering the extensive amount of data. Moreover,
ubiquitous computing is provided by the collaboration of several and different enti-
ties, creating new mobility solutions for communicating devices, which may become
targets to attackers. Together with Software-Defined Networking (SDN) and Network
Function Virtualization (NFV),EC has been deemed as a key enabling technology
123
toward the 5G era [89]. Besides, the integration between EC and 5G will grant better
computation and near real-time response, disregarding the physical location and the
Internet of Thing device resources. The vast amount of applications, especially the
location-based with high accuracy demand, increase the complexity of the deploy-
ment [32]. The mentioned 5G properties and features represent new implementation
challenges [66].
According to Hassan et al. [53], EC will have six important roles in 5G networks.
Local storage, computation, data analysis, decision-making, operation, and security
enhancement. Meaning that many features associated with Cloud services will be
provided by nearby edge nodes.
With 5G networks, data acquisition strategies, data storage methods, remote data
management, and decision-making will need to be severely faster, thus data pre-
processing will have to be upgraded. The development of Machine-to-Machine and
Machine-to-Infrastructure communication that 5G networks will enhance also pro-
vides new research opportunities in areas such as autonomous vehicles, drones, and
healthcare.
5.3 Mobility
The number of mobile devices connected to the network edge is growing at a fast
pace. While mobility adds flexibility to users and applications, it also brings critical
challenges. Mobility is responsible for many of the disconnections between edge
devices and the edge network, thus lowering the global quality of the service with a
high impact on parameters such as loss, delay, and bandwidth [2].
However, a global management architecture to enable the seamless handover of
devices and services, implementing fault-tolerance systems to reduce the impact of
failures, is yet to be presented. As different network access technologies and admin-
istrative domains may be available to the mobile device, both horizontal and vertical
handovers must be considered and optimized. This optimization must consider not
only the quality of the signal received but also other parameters, such as the direction
of the movement, the cost-benefit between different networks, or the quality of the
service available.
Moreover, it is important to consider the mobility of the edge nodes as it increases the
requirements of resource availability, resource discovery, task offloading, and resource
provisioning [125]. User mobility may also have a high impact on the number of
hops between a user and its services, mainly if this transition happens at the network
boundaries. Thus, such network changes compel quick and dynamic migration or
replication of edge services [43].
5.4 Security and privacy
In highly heterogeneous edge networks, security may have different requirements

depending on the devices and applications used. While some devices with lower
capabilities demand lightweight security, delay-sensitive networks, such as vehicular
123
networks, demand highly efficient security mechanisms to address their delay-sensitive

applications. As it sends sensitive data through the network, it raises privacy concerns.
The vulnerability to attacks in distributed edge environments, especially users’
devices that connect to edge nodes, is higher than in centralized systems because of
resource-restricted edge nodes and the lack of intelligence within to detect an attack.
Third-party stakeholders can draw dividends from accessing an edge platform, the
user’s information, and network analytics from it. Thus, contemplating authentication
based on privileges is necessary, despite the difficulties of intrusion detection in vast
geo-distributed elements [108].
EC architectures require uncompromising security methods fitting different access
technologies and devices, among services that have distinct security demands. For
example, in a network, various entities may manage that and where the coexistence of
multiple trust domains still lacks proper management and access control. Arising from
the context-awareness feature of EC, developing environment explicit security-aware
software systems is also a requirement. Furthermore, it is also needed to extend existing
trust models that focus on both users and network to include also the multiple service
providers, demanding new trust models, and identity management mechanisms.
5.5 Heterogeneity
The rising number and diversity of connected devices, with different hardware con-
straints, some mobile, together with a growing number of network access technologies
such as 3G, 4G, Wi-Fi, WiMAX, and the new 5G, which envisions the integration
of current devices with the IoT, contributes to a highly heterogeneous environment
in today’s edge systems. This heterogeneity not only implies an additional effort to
enable seamless handover between different network technologies but also to manage
several requirements from devices that demand unique resources from the network,
such as computing and storage, to overcome their constraints. A network and device
management that respects the timings required by the new applications optimizes the
communications and provides efficient use of devices and available edge resources is
a requisite to the success of the EC paradigm.
5.6 Reliability
EC may take advantage of the Cloud to increase reliability. By storing data or execute
applications in the Cloud servers, it diminishes the possibility of data and application
loss on mobile devices [30]. It is essential to deploy reliable edge systems without
the support of Cloud servers. Also, tackle some issues, such as failure of individ-
ual devices, network coverage, network failure, platform failure, and user interface
failure [81], a system that provides a reliable and fault-tolerant edge environment
is yet to be developed. Battery constraints, connection fluctuations, device availabil-
ity, and mobility problems also contribute to the difficulty of achieving such a system.
Also, checkpointing and rescheduling techniques used in centralized systems still lack
implementation in EC because of the mobility, heterogeneity, dynamic, and latency
requirements of these environments.
123
5.7 Storage data models
Managing storage features in all the edge devices, which are heterogeneous amongst
them, is a demanding task in edge-Cloud environments, considering the possibilities
of data ownership, data format, Application Programming Interfaces provided by data
owners, and computation platforms. A solution to this challenge would be the creation
of a unique data model for all layers of the system, fitting the before mentioned archi-
tectures. Naturally, its implementation is hard, and it must address several features,
such as device discovery, data exchange policies between nodes from different layers,
communication, and network protocols, automating data consistency, and summariza-
tion processes. This transversal data model would necessarily need to be automatic,
transparent to the user, enabling its availability to every level, regardless of the data,
its location, and its load. Also, to support this data model, a new framework, devel-
oped under a Platform as a Service (PaaS), would facilitate the development of edge
applications through a user-friendly solution.
5.8 Simulation environments
Being a recent concept, EC needs simulation environments to model a real-world

implementation. Thus, a simulation platform with a general programming language
(defining the language is, on its own, a significant challenge) would allow researchers
to configure, test, reconfigure and experiment with different scenarios and use cases.
Lowering the costs by simulating a proper model instead of its outdoor implementation
would be a critical benefit of such a platform, giving a holistic perception of the
system and its feasibility [2]. Also, simulation tools are useful to assess the benefits
of transforming a non-EC system into an EC system before its deployment [70].
There are already some valuable works on simulation environments. Sonmez et al.
[104] created EdgeCloudSim, a simulation mechanism based on CloudSim for perfor-
mance evaluation of EC systems. Kang et al. [64] implemented a simulation offloading
for 5G mobile terminal in Kubernetes. Mahmud and Buyya [73] introduced iFogSim,
which examines high-level steps to model and simulate Fog Computing environments.
Huang et al. [56] implemented computer simulation techniques for solving Stochas-
tic Petri Net models practically, to address reliability-aware service composition EC.
Jha et al. [60] introduced IoTSim-Edge a highly configurable platform, also based on
CloudSim, aiming to seize the behavior of heterogeneous IoT and EC infrastructures.
Lee et al. [68] introduced SIMaaS, a simulation as a service for Multi-access Edge
Computing, by offloading the simulation using a Cloud infrastructure based on the
concept of computation offloading.
Notwithstanding the development of these simulation tools, there are still some
limitations since each address only a slender number of features of the EC environ-
ment. Most approaches neglect the heterogeneity of devices, their battery constraints,
communication and network protocols, and emulation mechanisms. IoTSim-Edge [60]
is the most complete tool but still lacks the implementation of emulation mechanisms.
123
5.9 Billing
Cloud solutions are already well implemented and studied, thus enabling Cloud ser-
vices providers to define clear pricing policies. Pricing EC is far more complex. The
requirements for charging each customer depend on network usage information, which
telecommunication operators control, on data produced, processed, and stored that
involves other entities, and that differs with the system requirements, resulting in
non-standardized pricing for the services [7]. Tariff planning depends on the demand,
which means dynamic pricing models are essential. One solution may include eco-
nomic incentives for providing EC devices to other users, however, this business model
is still nonexistent [108].
5.10 Industry 4.0
Industry 4.0 (I4.0) integrates IoT, industrial internet, smart manufacturing [112], cloud
computing, intelligent robotics, and wireless sensor networks to implement cyber-
physical industrial systems [17]. EC is the next step in Industry 4.0 (I4.0) since it will
promote the flexibility, adaptation, and level of control that EC systems require for
fast-changing environments. Security and privacy concerns are also relevant topics in
Industry 4.0 (I4.0) since the Internet is the most commonly used means for transferring
data between different locations [13].
Edge networks also face many other challenges in these industrial environments. For
instance, it is imperative to address robustness and reliability issues in these extreme
and complex environments. High temperatures, dust, humidity, and electromagnetic
interference, among other problems, represent real challenges for the reliability of
I4.0 and EC architectures in particular. Also, equipment compatibility problems will
most probably impose more complex solutions that would be desirable to handle edge
nodes’ diversity. Many pieces of equipment will be analogic, others digital, others
mechanical, and require different monitoring, sensing, actuation, and communication
technologies. It is imperative to standardize the different data formats. Furthermore,
the advancements needed for EC in I4.0 cannot impair the more basic industries. Other
open issues are: improving computation offloading algorithms [17], developing new
architectures based on their needs [101], managing the volume and heterogeneity of
data produced, or performing real-time data analytics while maintaining data integrity
and redundancy [13].
The list of research opportunities in this section is not, by any means, a full list of
everything this area needs to achieve. From the current literature review standpoint,
we consider that it contains some of the most relevant problems to tackle in the future.
One of the key ideas to keep is that having a user-friendly and easy development of
applications is mandatory for the prosperity of the EC paradigm, so it is fundamental
to address these challenges in a transparent and holistic approach.
123
6 Conclusions
Edge Computing (EC) systems are still recent, and their development will result in de
facto computing architectures for everyday reality. With the existence of 5G, central-
ized methodologies will become obsolete when dealing with the amount of produced
and consumed data by the edge devices and end users. During the last decade, there
has been an increasing amount of research effort in this field to take advantage of
the potentials of these new architectures, replacing or complementing existing Cloud
systems.
This paper scrutinized the EC paradigm and its architectures, such as Multi-access
Edge Computing, Fog Computing, Cloudlet Computing, and Mobile Cloud Comput-
ing. This work attempted to give the reader useful insights about their definition and
essential properties, despite the difficulties to define boundaries between the mentioned
architectures. Also identified the most common use cases in EC and provided a pro-
found analysis of existing Smart Home/Building/City scenarios where several authors
deployed EC systems. This paper displayed research directions chronologically and
detailed some promising open challenges for future research: computation offload-
ing, 5G and EC, mobility, security and privacy, heterogeneity, reliability, storage data
models, simulation environments, billing, and Industry 4.0.
In conclusion, this work helps to clarify the scope of EC, while also provid-
ing state-of-the-art and relevant knowledge for EC researchers, industry developers,
entrepreneurs, and technology providers. As future work, we intend to work on some
proposed research challenges, namely Storage Data Models, expanding the existing
data models to the current needs of distributed systems.
Acknowledgements This work is supported by the European Regional Development Fund (FEDER),
through the Regional Operational Programme of Lisbon (POR LISBOA 2020) and the Competitiveness and
Internationalization Operational Programme (COMPETE 2020) of the Portugal 2020 framework [Project
5G with Nr. 024539 (POCI-01-0247-FEDER-024539)]. We also acknowledge the support from the Mobi-
Wise project: from mobile sensing to mobility advising (P2020 SAICTPAC/0011/2015), co-financed by
COMPETE 2020, Portugal 2020-POCI, European Regional Development Fund of European Union, and
the Portuguese Foundation of Science and Technology.
References
1. Abbas N, Zhang Y, Taherkordi A, Skeie T (2017) Mobile edge computing: a survey. IEEE Internet
Things J 5(1):450–465
2. Ahmed A, Ahmed E (2016) A survey on mobile edge computing. In: 10th international conference
on intelligent systems and control (ISCO’16). pp 1–8
3. Aldmour R, Yousef S, Yaghi M, Tapaswi S, Pattanaik KK, Cole M (2017) New cloud offloading
algorithm for better energy consumption and process time. Int J Syst Assur Eng Manag 8(s2):730–
733
4. Ayad M, Taher M, Salem A (2014) Real-time mobile cloud computing: a case study in face recognition.
In: 28th International conference on advanced information networking and applications workshops.
pp 73–78
5. Badidi E (2020) Qos-aware placement of tasks on a fog cluster in an edge computing environment. J
Ubiquitous Syst Pervasive Netw 13(1):11–19
6. Bagchi S, Siddiqui MB, Wood P, Zhang H (2020) Dependability in edge computing. Commun ACM
63(1):58–66
123
7. Baktayan A, AlGabri M, Alhomdy S (2018) Fog computing for network slicing in 5G networks: an
overview. J Telecom Syst Manag 07(02):1–18
8. Baktir AC, Ozgovde A, Ersoy C (2017) How can edge computing benefit from software-defined
networking: a survey, use cases, and future directions. IEEE Commun Surv Tutor 19(4):2359–2391
9. Barbarossa S, Sardellitti S, Di Lorenzo P (2014) Communicating while computing: distributed mobile
cloud computing over 5G heterogeneous networks. IEEE Signal Process Mag 31(6):45–55
10. Beck MT, Werner M, Feld S, Schimper T (2014) Mobile edge computing: a taxonomy. In: 6th
International conference on advances in future internet, (AFIN). pp 48–54
11. Bilal K, Khalid O, Erbad A, Khan SU (2018) Potentials, trends, and prospects in edge technologies:
fog, cloudlet, mobile edge, and micro data centers. Comput Netw 130:94–120
12. Billah F, Adnan M (2019) Smartlet: a dynamic architecture for real time face recognition in smartphone
using cloudlets and cloud. Big Data Res 17:45–55
13. Bodkhe U, Tanwar S, Parekh K, Khanpara P, Tyagi S, Kumar N, Alazab M (2020) Blockchain for
industry 4.0: a comprehensive review. IEEE Access 8:79764–79800
14. Bonomi F, Milito R, Natarajan P, Zhu J (2014) Fog computing: a platform for internet of things and
analytics. Big Data Internet Things Roadmap Smart Environ 546:169–186
15. Bonomi F, Milito R, Zhu J, Addepalli S (2012) Fog computing and its role in the internet of things.
In: Proceedings of the First Edition of the MCC workshop on mobile cloud computing. Association
for Computing Machinery, Helsinki, Finland, pp 13–16. https://doi.org/10.1145/2342509.2342513
16. Bou Abdo J, Demerjian J (2017) Evaluation of mobile cloud architectures. Pervasive Mobile Comput
39(December):284–303
17. Cao Z, Zhou P, Li R, Huang S, Wu D (2020) Multiagent deep reinforcement learning for joint
multichannel access and task offloading of mobile-edge computing in industry 4.0. IEEE Internet
Things J 7(7):6201–6213
18. Carvalho G, Cabral B, Pereira V, Bernardino J (2019) A case for machine learning in edge-oriented
computing to enhance mobility as a service. In: 15th International conference on distributed computing
in sensor systems, (DCOSS’19). pp 530–537
19. Chanakya B, Kiran PS (2017) A comprehensive survey of fog computing with internet of everything
(IoE). Int J Control Theory Appl 10(29):99–106
20. Chandavale A, Gade A, Dixit A (2019) Medical knowledge extraction scheme for cloudlet-based
healthcare system to avoid malicious attacks. Int J Cloud Comput 8(4):319–331
21. Chen L, Wu J, Zhou G, Ma L (2018) QUICK: qos-guaranteed efficient cloudlet placement in wireless
metropolitan area networks. J Supercomput 74(8):4037–4059
22. Chen N, Chen Y, You Y, Ling H, Liang P, Zimmermann R (2016) Dynamic urban surveillance video
stream processing using fog computing. In: Proceedings—016 IEEE 2nd international conference on
multimedia big data, BigMM 2016. pp 105–112
23. Chiang M, Ha S, I, CL, Risso, F, Zhang T, (2017) Clarifying fog computing and networking: 10
questions and answers. IEEE Commun Mag 55:18–20
24. Chiang M, Zhang T (2016) Fog and IoT: an overview of research opportunities. IEEE Internet Things
J 3(6):854–864
25. Consortium O (2017) OpenFog reference architecture for fog computing. Technical report
26. Dastjerdi A, Gupta H, Calheiros R, Ghosh S, Buyya R (2016) Chapter-4 fog computing: principles,
architectures, and applications. In: Internet of things. pp 61–75
27. Datla D, Chen X, Tsou T, Raghunandan S, Hasan SM, Reed JH, Dietrich CB, Bose T, Fette B, Kim
JH (2012) Wireless distributed computing: a survey of research challenges. IEEE Commun Mag
50(1):144–152
28. Davis A, Parikh J, Weihl WE (2004) Edgecomputing: extending enterprise applications to the edge
of the internet. In: Proceedings of the 13th international World Wide Web conference on Alternate
track papers and posters. pp 180–187
29. De D, Mukherjee A, Roy DG (2020) Power and delay efficient multilevel offloading strategies for
mobile cloud computing. Wirel Pers Commun 112(4):2159–2186
30. Dinh HT, Lee C, Niyato D, Wang P (2013) A survey of MCC: architecture, applications, and
approaches. Wirel Commun Mobile Comput 13:1587–1611
31. Dolezal J, Becvar Z, Zeman T (2016) Performance evaluation of computation offloading from mobile
device to the edge of mobile network. In: 2016 IEEE conference on standards for communications
and networking, CSCN 2016. pp 1–7
123
32. Duan Q, Wang S, Ansari N (2020) Convergence of networking and cloud/edge computing: status,
challenges, and opportunities. IEEE Netw 34:1–8
33. Dubey H, Yang J, Constant N, Amiri AM, Yang Q, Makodiya K (2015) Fog data: enhancing telehealth
big data through fog computing. In: ASE BigData and socialInformatics (ASE BD&SI). pp 1–6
34. El-Sayed H, Sankar S, Prasad M, Puthal D (2018) Edge of things: the big picture on the integration
of edge. IoT and the Cloud. IEEE Access 6:1706–1717
35. ETSI: MEC 003 - V2.1.1-Multi-access edge computing (MEC); framework and reference architecture.
Technical report (2019)
36. Fernández-CaramésTM Fraga-Lamas P, Suárez-Albela M, Vilar-Montesinos M (2018) A fog com-
puting and cloudlet based augmented reality system for the industry 4.0 shipyard. Sensors 18(6):1798
37. Fernando N, Loke SW, Rahayu W (2013) Mobile cloud computing: a survey. Future Gener Comput
Syst 29(1):84–106
38. Fernando N, Loke SW, Rahayu W (2016) Computing with nearby mobile devices: a work sharing
algorithm for mobile edge-clouds. IEEE Trans CC 7161:1–14
39. Ferrer AJ, Marquès JM, Jorba J (2019) Towards the decentralised cloud: survey on approaches and
challenges for mobile, ad-hoc and edge computing. ACM Comput Surv 51(6):1–39
40. Firdhous M, Ghazali O, Hassan S (2014) Fog computing: will it be the future of cloud computing?
In: 3rd International conference on informatics and applications. pp 8–15
41. Gao Z, Hao W, Zhang R, Yang S (2020) Markov decision process-based computation offloading
algorithm and resource allocation in time constraint for mobile cloud computing. IET Commun
14(13):2068–2078
42. Garcia Lopez P, Montresor A, Epema D, Datta A, Higashino T, Iamnitchi A, Barcellos M, Felber P,
Riviere E (2015) Edge-centric computing. ACM SIGCOMM Comput Commun Rev 45(5):37–42
43. Gedeon J, Brandherm F, Egert R, Grube T, Mühlhäuser M (2019) What the fog? edge computing
revisited: promises. Applications and future challenges. IEEE Access 7:152847–152878
44. Gedeon J, Krisztinkovics J, Meurisch C, Stein M, Wang L, Mühlhäuser M (2018) A multi-cloudlet
infrastructure for future smart cities: an empirical study. In: 1st International workshop on edge
systems, analytics and networking. pp 19–24
45. Giordano A, Spezzano G, Vinci A (2016) Smart agents and fog computing for smart city applications.
In: International conference smart cities. pp 137–146
46. Gonzalez NM, Goya WA, Silva EA, Cristina T, Brito MD (2016) Fog computing: data analytics and
cloud distributed processing on the network edges. In: 35th International conference of the Chilean
computer science society, (SCCC). pp 1–9
47. Grewe D, Wagner M, Arumaithurai M, Psaras I, Kutscher D (2017) Information-centric mobile
edge computing for connected vehicle environments. In: Workshop on mobile edge communications,
(MECOMM). pp 7–12
48. Gu Z, Takahashi R, Fukazawa Y (2019) Real-time resources allocation framework for multi-task
offloading in mobile cloud computing. In: International conference on computer, information and
telecomm, systems, CITS’19. pp 1–5
49. Guan T, Zaluska E, De Roure D (2005) A grid service infrastructure for mobile devices. In: 1st
international conference on semantics, knowledge and grid. pp 2–5
50. Gupta H, Chakraborty S, Ghosh SK, Buyya R (2016) Fog computing in 5G networks: an application
perspective. Fog 5G:1–36
51. Hall P, Miller H (2018) Fog computing architecture, evaluation, and future research directions. IEEE
Commun Mag 56:46–52
52. Han D, Chen W, Bai B, Fang Y (2019) Offloading optimization and bottleneck analysis for mobile
cloud computing. IEEE Trans Commun 67(9):6153–6167
53. Hassan N, Yau KLA, Wu C (2019) Edge computing in 5G: a review. IEEE Access Special Section
on MEC and MCC 7:127276–127289
54. Hong CH, Varghese B (2019) Resource management in fog/edge computing: a survey on architectures,
infrastructure, and algorithms. ACM Comput Surv 52(5):1–37
55. Hu P, Dhelim S, Ning H, Qiu T (2017) Survey on fog computing: architecture, key technologies,
applications and open issues. J Netw Comput Appl 98:27–42. https://doi.org/10.1016/j.jnca.2017.
09.002
56. Huang J, Liang J, Ali S (2020) A simulation-based optimization approach for reliability-aware service
composition in edge computing. IEEE Access 8:50355–50366
123
57. Issarny V, Georgantas N, Hachem S, Zarras A, Vassiliadist P, Autili M, Gerosa MA, Hamida AB
(2011) Service-oriented middleware for the Future Internet: state of the art and research directions. J
Internet Serv Appl 2(1):23–45
58. Jararweh Y, Doulat A, Alqudah O, Ahmed E, Al-Ayyoub M, Benkhelifa E (2016) The future of mobile
cloud computing: integrating cloudlets and mobile edge computing. In: 23rd International conference
on telecommunications, (ICT). pp 1–5
59. Javadzadeh G, Rahmani AM (2020) Fog computing applications in smart cities: a systematic survey.
Wireless Netw 26(2):1433–1457
60. Jha D, Alwasel K, Alshoshan A, Huang X, Naha R, Battula S, Garg S, Puthal D, James P, Zomaya
A, Dustdar S, Ranjan R (2020) IoTSim-Edge: a simulation framework for modeling the behavior of
IoT and EC environments. Softw Pract Exp 50:1–19
61. Jia G, Han G, Li A, Du J (2018) SSL: smart street lamp based on fog computing for smarter cities.
IEEE Trans Ind Inf 14(11):4995–5004
62. Jia M, Liang W, Xu Z (2017) Qos-aware task offloading in distributed cloudlets with virtual network
function services. In: Proceedings of the 20th ACM international conference on modelling, analysis
and simulation of wireless and mobile systems, pp 106–119
63. Jiang C, Cheng X, Gao H, Zhou X, Wan J (2019) Toward computation offloading in edge computing:
a survey. IEEE Access 7:131543–131558
64. Kang S, Lee J, Jeon J, Chun I (2019) Multi-access edge computing based simulation offloading for
5g mobile application. In: 17th annual international conference on mobile systems, applications, and
services. pp 590–591
65. Khan WZ, Ahmed E, Hakak S, Yaqoob I, Ahmed A (2019) Edge computing: a survey. Future Gener
Comput Syst 97:219–235
66. Kiss P, Reale A, Ferrari CJ, Istenes Z (2018) Deployment of IoT applications on 5G edge. In: IEEE
international conference on future IoT technologies. pp 1–9
67. Kitanov S, Monteiro E, Janevski T (2016) 5G and the fog-survey of related technologies and research
directions. In: 18th Mediterranean Electrotechnical conference: intelligent and efficient technologies
and services for the citizen. pp 18–20
68. Lee J, Kang S, Jeon J, Chun I (2020) Multiaccess edge computing-based simulation as a service for
5G mobile applications: a case study of tollgate selection for autonomous vehicles. Wirel Commun
Mobile Comput. https://doi.org/10.1155/2020/9869434
69. Li C, Xue Y, Wang J, Zhang W, Li T (2018) Edge-oriented computing paradigms: a survey on
architecture design and system management. ACM Comput Surv 51(2):A34–A39
70. Liu F, Tang G, Li Y, Cai Z, Zhang X, Zhou T (2019) A survey on edge computing systems and tools.
Proc IEEE 107(8):1537–1562
71. Luan TH, Gao L, Li Z, Xiang Y, Wei G, Sun L Comput Sci 1–11
72. Mach P, Becvar Z (2017) Mobile edge computing: a survey on architecture and computation offload-
ing. IEEE Commun Surv Tutor 19(3):1628–1656
73. Mahmud R, Buyya R (2019) Fog and edge comp: principles and paradigms, 1st edn
74. Mahmud R, Kotagiri R, Buyya R (2016) Fog computing: a taxonomy, survey and future directions.
pp 1–28
75. Mazza D, Tarchi D, Corazza GE (2017) A unified urban mobile cloud computing offloading mecha-
nism for smart cities. IEEE Commun Mag 55(3):30–37
76. Mehta S, Kaur P (2019) Efficient computation offloading in mobile cloud computing with nature-
inspired algorithms. Int J Comput Intell Appl 18(4):1950023
77. Mouradian C, Naboulsi D, Yangui S, Glitho RH, Morrow MJ, Polakos PA (2018) A comprehen-
sive survey on fog computing: state-of-the-art and research challenges. IEEE Commun Surv Tutor
20(1):416–464
78. Muniswamaiah M, Tappert CC (2019) Mobile cloud computing in healthcare using dynamic cloudlets
for energy-aware consumption. CoRR abs/1908.11501
79. Naha RK, Garg S, Georgakopoulos D, Jayaraman PP, Gao L, Xiang Y, Ranjan R (2018) Fog
computing: survey of trends, architectures, requirements, and research directions. IEEE Access
6:47980–48009
80. Nastic S, Rausch T, Scekic O, Dustdar S, Gusev M, Koteska B, Kostoska M, Jakimovski B, Ristov S,
Prodan R (2017) A serverless real-time for edge computing. IEEE Internet Comput Internet 21:64–71
81. Nath SB, Gupta H, Chakraborty S, Ghosh SK (2018) A survey of fog computing and communication:
current researches and future directions. IEEE Access (i) 1–47
123
82. Ning H, Li Y, Shi F, Yang LT (2020) Heterogeneous edge computing open platforms and tools for
internet of things. Future Gener Comput Syst 106:67–76
83. Noor TH, Zeadally S, Alfazi A, Sheng QZ (2018) Mobile cloud computing: challenges and future
research directions. J Net Comput Appl 115:70–85
84. Nunna S, Kousaridas A, Ibrahim M, Dillinger M, Thuemmler C, Feussner H, Schneider A (2015)
Enabling real-time context-aware collaboration through 5G and MEC. In: 12th international confer-
ence on information technology: new generations. pp 601–605
85. Pang Z, Sun L, Wang Z, Tian E, Yang S (2016) A survey of cloudlet based mobile computing. In:
international conference on cloud computing and big data. pp 268–275
86. Patel M, Hu Y, Hédé P, Joubert J, Thornton C, Naughton B, Julian RR, Chan C, Young V, Tan SJ, Lynch
D (2014) Mobile edge computing-introductory technical white paper. ETSI White Paper 11(1):1–36
87. Rahimi MR, Ren J, Liu CH, Vasilakos AV, Venkatasubramanian N (2014) Mobile cloud computing:
a survey, state of art and future directions. Mobile Netwo Appl 19(2):133–143
88. Ray PP, Dash D, De D (2019) Edge computing for internet of things: a survey, e-healthcare case study
and future direction. J Net Comput Appl 140:1–22
89. Ren J, Zhang D, He S, Zhang Y, Li T (2019) A survey on end-edge-cloud orchestrated network
computing paradigms: transparent computing, mobile edge computing, fog computing, and cloudlet.
ACM Comput Surv 52(6):1–36
90. Roman R, Lopez J, Mambo M (2018) Mobile edge computing, Fog et al.: a survey and analysis of
security threats and challenges. Future Gener Comput Syst 78:680–698
91. Sabella D, Vaillant A, Kuure P, Rauschenbach U, Giust F (2016) Mobile-edge computing architecture:
the role of MEC in the internet of things. IEEE Consum Electron Mag 5(4):84–91
92. Sangal SMHKVAL (2015) Analysis of cloudlet completion time during attack on smart grid cloud.
Int J Cloud Comput 4:356–376
93. Satyanarayanan M (2017) The emergence of edge computing. Computer 50(1):30–39
94. Satyanarayanan M, Bahl P, Cáceres R, Davies N (2009) The case for VM-based cloudlets in mobile
computing. IEEE Pervasive Comput 8(4):14–23
95. Shahzadi S, Iqbal M, Dagiuklas T, Qayyum ZU (2017) Multi-access edge computing: open issues,
challenges and future perspectives. J Cloud Comput 6(1):30
96. Shi W, Cao J, Zhang Q, Li Y, Xu L (2016) Edge computing: vision and challenges. IEEE Internet
Things J 3(5):637–646
97. Shi W, Dustdar S (2016) The promise of edge computing. Computer 49:78–81
98. Simoens P, Xiao Y, Pillai P, Chen Z, Ha K, Satyanarayanan M (2013) Scalable crowd-sourcing of
video from mobile devices. In: 11th annual international conference on mobile systems, applications,
and services, (MobiSys ’13). p 139
99. Sinaeepourfard A, Krogstie J, Petersen SA, Ahlers D (2019) F2c2c-dm: a fog-to-cloudlet-to-cloud
data management architecture in smart city. In: 2019 IEEE 5th world forum on internet of things
(WF-IoT). pp 590–595
100. Sinky H, Hamdaoui B (2016) Cloudlet-aware mobile content delivery in wireless urban communica-
tion networks. In: 2016 IEEE global communications conference, GLOBECOM 2016, Washington,
DC, USA, December 4–8, 2016, IEEE. pp 1–7
101. Sittón-Candanedo I, Alonso R, Rodríguez-González S, Coria J, de la Prieta F (2019) Edge computing
architectures in industry 4.0: a general survey and comparison. In: 14th International conference
on soft computing models in industrial and environmental applications (SOCO 2019), vol 950. pp
121–131
102. Sneps-Sneppe M, Namiot D (2016) On mobile cloud for smart city applications. CoRR
103. Song Y, Yau SS, Yu R, Zhang X, Xue G (2017) An approach to qos-based task distribution in edge
computing networks for iot applications. In: IEEE international conference on edge computing. IEEE
Computer Society, pp 32–39
104. Sonmez C, Ozgovde, A, Ersoy, C (2017) EdgeCloudSim: an environment for performance evaluation
of edge computing systems. In: 2nd International conference on fog and mobile edge computing,
(FMEC’17). pp 39–44
105. Stojmenovic I, Wen S (2014) The fog computing paradigm: scenarios and security issues. In: Federated
conference on computer science and information systems, vol 2. pp 1–8
106. Sun C, Li H, Li X, Wen J, Xiong Q, Zhou W (2020) Convergence of recommender systems and EC:
a comprehensive survey. IEEE Access 8:47118–47132
123
107. Sun X, Ansari N (2017) Latency aware workload offloading in the cloudlet network. IEEE Commun
Lett 21(7):1481–1484
108. Taleb T, Samdanis K, Mada B, Flinck H, Dutta S, Sabella D (2017) On multi-access edge computing:
a survey of the emerging 5G network edge cloud architecture and orchestration. IEEE Commun Surv
Tutor 19(3):1657–1681
109. Tawalbeh LA, Bakheder W, Mehmood R, Song H (2016) Cloudlet-based mobile cloud computing
for healthcare applications. In: IEEE global communications conference, (GLOBECOM). pp 1–6
110. Tran TX, Hajisami A, Pandey P, Pompili D (2017) Collaborative mobile edge computing in 5G
networks: new paradigms, scenarios, and challenges. IEEE Commun Mag 55(4):54–61
111. Tuli S, Basumatary N, Gill SS, Kahani M, Arya RC, Wander GS, Buyya R (2020) HealthFog: an
ensemble deep learning based smart healthcare system for automatic diagnosis of heart diseases in
integrated IoT and fog computing environments. Future Gener Comput Syst 104:187–200
112. Vaidya S, Ambad P, Bhosle S (2018) Industry 4.0–a Glimpse. Procedia Manuf 20:233–238
113. Vaquero LM, Rodero-Merino L (2014) Finding your way in the fog. ACM SIGCOMM Comput
Commun Rev 44(5):27–32
114. Varshney P, Simmhan Y (2017) Demystifying fog computing: characterizing architectures, applica-
tions and abstractions. In: IEEE 1st International conference on fog and edge computing (ICFEC’17).
pp 115–124
115. Wang S, Zhang X, Zhang Y, Wang L, Yang J, Wang W (2017) A survey on mobile edge networks:
convergence of computing, caching and communications. IEEE Access SS Secur Anal Intell CPS
5:6757–6779
116. Wang T, Luo H, Zheng X, Xie M (2019) Crowdsourcing mechanism for trust evaluation in CPCS
based on intelligent mobile edge computing. ACM Trans Intell Syst Technol 10(6):62:1–62:19
117. Wang Y, Chen IR, Wang DC (2015) A survey of mobile cloud computing applications: perspectives
and challenges. Wirel Pers Commun 80(4):1607–1623
118. Wang Y, Pan Y (2015) Cloud-dew architecture: realizing the potential of distributed database systems
in unreliable networks. In: Proceedings of the international conference on parallel and distributed
processing techniques and applications (PDPTA). p 85
119. Yang B, Chai WK, Pavlou G, Katsaros KV (2016) Seamless support of low latency mobile applications
with NFV-enabled mobile edge-cloud. In: 5th IEEE international conference on cloud networking,
(CloudNet). pp 136–141
120. Yao D, Yu C, Yang LT, Jin H (2019) Using crowdsourcing to provide qos for mobile cloud computing.
IEEE Trans Cloud Comput 7(2):344–356
121. Yassine A, Hossain MS, Muhammad G, Guizani M (2020) Cloudlet-based intelligent auctioning
agents for truthful autonomous electric vehicles energy crowdsourcing. IEEE Trans Veh Technol
69(5):5457–5466
122. Yi S, Hao Z, Qin Z, Li Q (2016) Fog computing: platform and applications. In: 3rd Workshop on hot
topics in web systems and technologies. pp 73–78
123. Yi S, Li C, Li Q (2015) A survey of fog computing: concepts, applications and issues. In: Workshop
on mobile big data-mobidata ’15. pp 37–42
124. Yogi MK, Chandrasekhar K, Kumar GV (2017) Mist computing: principles, trends and future direc-
tion. SSRG Int J Comput Sci Eng 4(7):19–21
125. Yousefpour A, Fung C, Nguyen T, Kadiyala K, Jalali F, Niakanlahiji A, Kong J, Jue JP (2019) All
one needs to know about fog computing and related edge computing paradigms: a complete survey.
J Syst Architect 98:289–330
126. Yu J, Lee N, Pyo CS, Lee YS (2018) WISE: web of object architecture on IoT environment for smart
home and building energy management. J Supercomput 74(9):4403–4418
127. Zhang J, Chen B, Zhao Y, Cheng X, Hu F (2018) Data security and privacy-preserving in edge
computing paradigm: survey and open issues. IEEE Access 6:18209–18237
128. Zhang J, Zhou Z, Li S, Gan L, Zhang X, Qi L, Xu X, Dou W (2018) Hybrid computation offloading
for smart home automation in mobile cloud computing. Pers Ubiquitous Comput 22(1):121–134
129. Zhang K, Mao Y, Leng S, He Y, Zhang Y (2017) Mobile-edge computing for vehicular networks.
IEEE Veh Technol Mag 12:36–44
130. Zhang Y (2004) Transparence computing: concept, architecture and example. Chin J Electron
32(12):169–174
131. Zhang Y, Niyato D, Wang P (2015) Offloading in mobile cloudlet systems with intermittent connec-
tivity. IEEE Trans Mob Comput 14(12):2516–2529
123
132. Zhuang W, Jamalipour A, Bai F, Vinel A (2017) Emerging technologies, applications, and standard-
izations for connecting vehicles. IEEE Veh Technol Mag 12(2):23–25
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.
123

Carvalho 2021

Uploaded by

Copyright:

Available Formats

Carvalho 2021

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Carvalho 2021

Uploaded by

Copyright:

Available Formats

Computing (2021) 103:993–1023

Edge computing: current trends, research challenges and

Gonçalo Carvalho1 · Bruno Cabral1 · Vasco Pereira1 · Jorge Bernardino1,2

Keywords Edge computing · Fog computing · Cloudlet computing · Multi-access

Mathematics Subject Classification 68-02

1 Department of Informatics Engineering, CISUC, University of Coimbra, Coimbra, Portugal

As the number of network-connected devices continues to increase at a fast pace,

2 The edge computing paradigm

Fig. 1 Edge computing deployment scenarios

Characteristics Cloud Edge

Applications Data-driven User-driven

3 Edge computing architectures

Multi-access Edge Computing: formerly known as Mobile Edge Computing,

Cloudlet Computing: the cloudlet concept was first proposed by Satyanarayanan

Mobile Cloud Computing: mobile devices are increasingly becoming an essen-

Table 2 Properties of edge computing architectures

Multi-access edge Fog computing Cloudlet Mobile cloud

Ownership Telecommunication Private entities or individuals

Multi-access edge Fog computing Cloudlet Mobile cloud

Power Low Low Low Low (mobile),

4 Edge computing use cases

Multi-access Edge Computing The scenarios that take advantage of Multi-access

Use case MEC Fog Cloudlet MCC

Use case MEC Fog Cloudlet MCC

Smart home/ building [1,96] [25,50,55,59, [45,79,122] [128]

Computing should contemplate the application of Cloudlet Computing for a higher

Multi-access edge computing Narrow to wide scope Expensive servers

5 Future research directions

5.1 Computation offloading

If an application requires high computational processing, or if the device running

Table 5 Chronologic evolution of edge computing research topics

Year EC research topics Papers

2011 Scalabihtv [57]

Year EC research topics Papers

applied to the computation offloading decision, it should answer questions such as

Year EC research topics Papers

Hardware technologies [69]

5.2 Edge computing and 5G

5.4 Security and privacy

In highly heterogeneous edge networks, security may have different requirements

networks, demand highly efficient security mechanisms to address their delay-sensitive

5.7 Storage data models

5.8 Simulation environments

Being a recent concept, EC needs simulation environments to model a real-world

5.10 Industry 4.0

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.