See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/329391382

Information-Centric Networking With Edge Computing for IoT: Research

Challenges and Future Directions

Article in IEEE Access · December 2018

DOI: 10.1109/ACCESS.2018.2884536

CITATIONS READS

68 4,937

3 authors:

Rehmat Ullah Syed Hassan Ahmed

Cardiff Metropolitan University JMA Wireless
67 PUBLICATIONS 1,032 CITATIONS 191 PUBLICATIONS 5,162 CITATIONS

SEE PROFILE SEE PROFILE

Byung-Seo Kim
Hongik University at Sejong Campus
132 PUBLICATIONS 2,016 CITATIONS

SEE PROFILE

All content following this page was uploaded by Rehmat Ullah on 18 December 2018.

The user has requested enhancement of the downloaded file.

SPECIAL SECTION ON SMART CACHING, COMMUNICATIONS, COMPUTING AND
CYBERSECURITY FOR INFORMATION-CENTRIC INTERNET OF THINGS

Received November 20, 2018, accepted November 28, 2018, date of publication December 3, 2018,
date of current version December 27, 2018.
Digital Object Identifier 10.1109/ACCESS.2018.2884536

Information-Centric Networking With Edge

Computing for IoT: Research Challenges
and Future Directions
REHMAT ULLAH1 , SYED HASSAN AHMED 2 , (Senior Member, IEEE),
AND BYUNG-SEO KIM 3 , (Senior Member, IEEE)
1 Department of Electronics and Computer Engineering, Hongik University, Sejong 30016, South Korea
2 Department of Computer Science, Georgia Southern University, Statesboro Campus, Statesboro, GA 30460, USA
3 Department of Software and Communications Engineering, Hongik University, Sejong 30016, South Korea
Corresponding author: Byung-Seo Kim ([email protected])
This work was supported by the National Research Foundation of Korea (NRF) funded by the Korea Government under
Grant 2018R1A2B6002399.

ABSTRACT Cloud computing is a paradigm that offers storage, computation, and software services on
demand in the Cloud and far from end users. These services and computations are then extended to the Edge
of the network with the Edge computing paradigm. This paradigm offers computation, data, and application
services in close proximity of end users. Future Internet architectures will result in a fast information
response time, and low latency will be a main feature of evolving 5th generation (5G) radio networks.
To ensure the widespread adoption of 5G applications, low latency, security, mobility, and scalable content
distribution support is required. Information-centric networking (ICN) is a newly proposed future Internet
paradigm in which communication is based on content names irrespective of their locations. At the same
time, ICN promises efficient content delivery, mobility support, scalability, and security for content. The
Edge computing and ICN provide an opportunity to reduce latency, support mobility, security, and scalability.
In this paper, discussions on Edge computing with ICN are provided. In detail, the Edge computing proposals,
use cases, differences among Edge computing proposals and drivers for Edge computing are investigated.
The Edge computing standardization, research, and industry/vendors collaboration overview are studied.
Applications of ICN integration in Edge computing and their advantages and limitations are highlighted.
We conclude our paper by describing potential directions for future research in the field of ICN over Edge
computing.

INDEX TERMS Internet of things, Information centric networking, cloud computing, edge computing,
caching, naming, ICN-IoT security services, self-certification, location-awareness, offloading, latency.

I. INTRODUCTION 3) Infra Structure as a Service (IaaS), which offers

The revolution of mobile services and Cloud Computing storage and computation services via virtualization
(CC) has recently attracted significant attention from both using frameworks, such as Amazon Elastic Compute
industry and academia. CC provides processing and stor- Cloud (AC2).
age for data in the cloud rather than inside devices of CC becomes problematic for applications that are latency
subscribers [1]. sensitive and for most of the requirements of novel used cases,
CC offers three main service types as shown in Figure 1: CC fall short. To overcome such limitations, an Edge com-
1) Software as a Service (SaaS), which offers access to puting paradigm has been introduced. The Edge computing
applications for end users, such as Facebook, Gmail, is an abstraction level paradigm that covers many related
and Microsoft Office 365; proposals with same principal and different aspects. The term
2) Platform as a Service (PaaS), which offers applica- Edge computing refers to the processing, storage and network
tion programming interface (API) development envi- optimization at the Edge of networks (fixed and mobile). The
ronments for developers, such as Microsoft Azure and Edge could be in the Radio access network (RAN) or in the
Amazon Web Services (AWS); and customer’s premises, or at some central location. The journey
2169-3536 2018 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 6, 2018 Personal use is also permitted, but republication/redistribution requires IEEE permission. 73465
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

and fixed purpose-built servers deployed at the network

edges [3].
Edge computing is a one of the trending topics which needs
further heed and affection. There are different deployment
strategies to boost the Edge computing performances such as
Software Defined Networking (SDN) and ICN. Edge Com-
puting paradigm is not directly related with the SDN imple-
mentations. However, it is noted that SDN has the capability
to overcome many challenges of Edge computing through
the flexibility of programmable networks [4]. SDN make
use of Network Functions Virtualization (NFV) technology
which separates the network functions from the hardware
and hand them over to a software-based application [5].
In order to enable SDN over Edge computing, related studies
should be examined by focusing on the requirements of edge
servers and functionalities of SDN. Although SDN is one
FIGURE 1. Cloud computing services. of the enabling technologies for Edge computing, however,
SDN with Edge computing is outside the scope of the present
study. Interested readers are referred to [6] for latest research
of Edge computing started due to the rise of Internet of things on the proposals for SDN over Edge Computing.
(IoT) and wearable devices. A huge amount of raw data may The Edge servers can be better managed through other
be generated in IoT and this may have different requirements. paradigms, such as ICN integration in Edge computing.
Therefore, to overcome such challenges, different groups In this article, we are focusing specifically on Named
come up with several proposals under the same paradigm Data Networking (NDN) over Edge Computing. However,
of Edge computing. However, CC and Edge computing are throughout this article we are using ICN as a general term.
not independent and have strong correlation. As a result, ICN is a promising future Internet paradigm solving
the paradigm of CC cannot be avoided completely. the information-dissemination problem through the naming
Although IoT is advantageous for many applications, how- approach. The content is retrieved using the name of the
ever it creates many challenges that cannot be addressed content, not the IP address as in the conventional Internet.
by today’s Cloud only model. Industrial IoT (IIoT) such We hope that ICN and Edge Computing would be one of the
as manufacturing systems, smart grids (to name a few), most important paradigms for 5G mobile networks. It will
often demand end-to-end latencies of a few milliseconds help to reduce the traffic load and latency and will also con-
between the sensor and the control node. Furthermore, other tribute to energy efficiency in green 5G networks. However,
IoT applications, such as vehicle-to-vehicle communications, there are several challenges that may result from the combina-
drone flight control applications, virtual reality applications, tion of ICN and Edge Computing, which are discussed further
and gaming applications, may require latencies below a few in this article. Moreover, we studied the Edge computing
tens of milliseconds. Beside latency the growing number of technologies in depth with supporting use case scenarios.
connected things is generating more data and thus creating As a summary, the contributions of this paper are discussed
bandwidth issues. Therefore, sending all the data to the Cloud as follows. We:
will require high network bandwidth. Moreover, mostly IoT 1) First provide a comprehensive review on the gen-
devices have resource constrained nature. These devices with eral concept of Cloud computing and Edge computing
limited resources will not be able to fulfill all their computing including the detailed reasons and motivation behind
needs using their own resources. Therefore, they need to the Edge computing paradigm
offload their computing needs directly to the Cloud which is 2) Survey the Edge computing architectures in detail with
unrealistic, because such interactions often require resource- targeted use cases (i-e what kind of use-cases can ben-
intensive processing and complex protocols. At the same efit from Edge computing), differences among various
time, the increasing number of connected devices also creates Edge computing proposals along with limitations and
a security challenge that how to update the security updates importance, and drivers for Edge computing
on each device. It is thus impractical to connect every device 3) Survey the Edge computing standardization activities,
to the Cloud in order to update its security credentials [2]. the involvement of industry and standard organizations
Edge computing can provide effective ways to overcome for addressing various research topics in Edge network
many limitations of the Cloud only model by distribut- and specifically ICN
ing computing, control, storage, and networking functions 4) Discuss various vendors which are working on the
closer to end user devices. However, the traditional model Edge computing solutions and up to date hardware
(IP approach) is still the dominant solution, where the end-to- and software solution related to the Edge computing
end communication is managed between the IoT data sources paradigm

73466 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

5) Introduce ICN as one of the enabling technologies

for Edge computing and illustrated how the intrinsic
properties of ICN has the potential to support and fulfill
the requirements (i-e Mobility, latency and security,
heterogeneity etc.) of future networks such as 5G and
beyond
6) Discuss recent works and ideas about the integration of
ICN and Edge computing
7) Discuss the limitations of both Edge computing and
ICN and highlighted the importance of leveraging ICN
FIGURE 2. Edge computing feature, approaches and goals.
over Edge Computing
8) Highlight various 5G applications that could be
improved by ICN over Edge computing
9) Finally provide future research directions for ICN A. WHY EDGE COMPUTING?
over Edge computing that could enhance the solution In section 1, we explained the Edge computing paradigm that
space. how conventional CC may fall short for fulfilling the require-
The paper is organized as follows. We present in depth ments of novel use cases. CC is centralized and not com-
the Edge computing paradigm in the Section 2. In section 3 patible with the diverse type of traffic generated by different
we present different proposals of Edge computing in detail. Edge devices. Hence, bringing servers to the Edge of network
In Section 4 we present the difference among all Edge becomes necessary. The necessity comes from various factors
computing proposals. Section 5 consisting of Edge com- that drive the evolution of Edge computing paradigm. These
puting novel use cases. Section 6 comprised of Edge com- factors need to be analyzed as view point of user and operator.
puting limitations. In Section 7 we present an overview of Figure 2. illustrates the features, approaches and goals of
ICN paradigm, advantages of leveraging ICN over Edge Edge computing. In the subsequent subsections, we discuss
computing and applications of ICN in Edge computing. the motivation behind Edge computing paradigm as follows:
Standardization activities and vendor’s solution related to
Edge computing and ICN are discussed in Section 8 and 1) QoS AND LATENCY
Section 9 respectively. Finally, we conclude the paper in Even though Edge devices are powerful, most of them are
section 10 and 11, by describing potential benefits and lacking enough capacity for fulfilling the delay sensitive
research directions for future research in the area of ICN over requirements. CC provides resource enriched technology
Edge computing. infrastructure for constrained devices with huge computa-
tion and storage capacity. However, most of the devices
II. EDGE COMPUTING in IoT i.e. wearable devices are delay sensitive. In legacy
Due to the evolution of cloud system, most of the functional- cloud paradigm, the accessibility of all these devices is done
ities were pushed to the centralized clouds. In this process of via Wide Area Network (WAN), which creates delay, and
virtualization where we got resource enrich platform, we also hence conventional cloud cannot deal with the problems such
lost the importance of the location and efficient utilization of as the mobility, and real-time requirements. Such latency-
resources. It is very true that physical distance increases the sensitive applications demanding computation power and
latency. The centralized architecture may be easy to manage memory resources that cannot be built satisfactorily using
and handle, however, may not fulfil the requirements of all cloud services, which can be many network hops away
end users. Many factors can affect the performances such from user locations. Use cases of IoT such as healthcare,
as traffic type, network condition, end user’s interests and needs high QoS requirements [2]. Computation resources
preferences. Therefore, Edge computing paradigm has been is required at the Edge of the network to meet high QoS.
proposed to address such challenges. As an example, in autonomous vehicles the generated data
The term Edge refers to the computing infrastructure that of camera need to be processed instantly to meet the real-
exists closer to the sources of data. These devices typi- time requirements of QoS [3]. User experience is affected by
cally reside away from the centralize computing available the centralized servers of cloud due to limited Internet band-
in the cloud. Edge computing pushes computing power to width and WAN delay. The overall latency can be reduced
the Edges of network, instead of centralized cloud. In Edge if the servers is deployed closer to user devices [4]. The
computing, the data analytics happens very close to the benefit of servers closer to the users is the high local-area
devices and sensors. Therefore, Edge computing thus results network (LAN) bandwidth and a smaller number of hopes.
in lower delay and high speed of task execution. More- An offloading scheme for IoT devices proposed in [5] shows
over, there is need for strong consensus between the ven- that utilizing servers at the Edge for processing the IoT data
dors and operators to keep a balance between what should provides a high reduction in the latency compared to the
be centralized and what should be at the Edge of the cloud servers. Moreover, Edge computing provides caching,
network [1]. storage and computation capabilities in close proximity of

VOLUME 6, 2018 73467

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

TABLE 1. Terms and abbreviations.

end users/devices thereby reduces end to end latency. Edge 3) SCALABILITY

computing offers such benefits without requiring deploy- It is predicated that the end user devices will reach tril-
ment in the core and remove cloud dependency for content lion and beyond that may create a serious scalability
processing. challenge [13], [14]. Therefore, sending huge amount of
data to the centralized servers create a congestion within the
2) MINIMIZATION OF CORE NETWORK TRAFFIC datacenters [15]. As a result, CC (centralized) may fall short
In the conventional CC approach, all the traffic flows from in the context of scalability for applications and data. The
devices through core network to reach cloud servers. The virtualization of Edge servers can bring an opportunity to
content as well as the context require processing and storage support scalability [16]. If any of the Edge server becomes
which may not achievable on mobile devices. Therefore, it is congested, then the request can be distributed to other Edge
done on the cloud that result in higher response time and servers in close proximity and so on. The burden on the cloud
increased backhaul traffic. In case of IoT billions of devices servers can be reduced by processing data at the Edge servers,
may generate a huge amount of raw data to be processed since smaller amount of traffic will be forwarded to the cloud
and stored. According to [11], 15 petabytes of traffic has servers [17].
been generated per month. Sending all the traffic to the In the subsequent section, we discuss the Edge computing
cloud servers may result in congestion on cloud servers, since drivers with features and goals.
cloud servers have limited capacity. To optimize bandwidth
utilization and to reduce traffic on the core network, the traffic
B. DRIVERS FOR EDGE COMPUTING
should be handled at the Edge servers [12]. Traffic from
billions of devices can be handled on Edge servers to prevent Centralized topologies fall short and are unable to serve
congestion and latency problems. Processing data at Edge traffic loads with Quality of Experience (QoE) for operators
servers also reduce the demand of computational resources and subscribers. We will discuss some of the key drivers for
at the cloud data centers such as processing of data generated Edge computing as follows:
from IoT sensors and cameras [13]. Therefore, Edge comput- 1) Continued growth in the user’s traffic such as video and
ing paradigm can play a meaningful role in traffic reduction interactive applications i-e. games becomes an essen-
on the core network. tial driver for Edge computing. Virtual reality (VR)

73468 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

FIGURE 3. Edge computing proposals.

requires increase capacity and lower latency. Edge some specific Edge location near to the subscribers. Appli-
computing will be useful in reducing the latency which cations with highly fluctuating network loads, a mobile Edge
is introduced by the backhaul. Physical distance of will definitely maximize the performance. However, shifting
backhaul result in high latency. To maximize the uti- Edge location according to the traffic characteristic is quite
lization of existing resources, Edge computing can play difficult to achieve [18].
a major role [18]. We explained the Edge computing paradigm so far. Con-
2) Using Edge computing the operators would move all sequently, we will explain different proposals related to the
the centralized hardware near to the end user(s). This Edge computing paradigm. A taxonomy of Edge computing
would result in the reduction of latency and efficient is shown in Figure 3.
utilization of bandwidth [18].
3) The benefits such as lower latency, location awareness, III. EDGE COMPUTING PROPOSALS
security and reduction in the core network traffic can be In this section, we will discuss in detail about various pro-
achieved if the Edge location is chosen appropriately posals which comes under the umbrella of Edge comput-
and right content is stored at the Edge. In case these ing paradigm. Initially we will discuss about Mobile Cloud
conditions are not met, the cost and complication may Computing (MCC), since it is the first initiatives in Edge
be increased with high price. computing, and then consequently we will discuss other Edge
4) Traffic is increasing day by day specifically video traf- computing proposals.
fic than other traffic. According to Cisco, video traffic
is 60% today, and would be 78% by 2021 [19]. Accord- A. MOBILE CLOUD COMPUTING
ing to estimates of Ericsson, the video traffic was 50% Due to proliferation of the mobile devices, mobile com-
in 2016, and would likely increase to 75% in 2022 [20]. puting terminologies have appeared to provide services to
Video has become an integral component of social mobile users. Mobile devices have limited resources and
media. The enormous increase and use of video traffic cannot provide demanded benefits and services to users.
give an important role to latency to showcase the user Nowadays, while mobile devices are rich in resources, they
experience. Therefore, the video traffic has been one of cannot still reach to the level of servers [21], [27]. Due to
the key drivers for Edge computing resource-constrained nature, mobile devices need to offload
tasks to cloud data centers for empowering various oper-
C. IMPORTANCE OF EDGE LOCATION ations. MCC offers offloading mechanism for the mobile
The most crucial question so far is: where exactly the location devices by integrating mobile Internet, mobile computing,
of Edge is.? From the Edge, we are meaning that process- and CC into a combined system [21]. CC is a service model
ing, storage, and control units move to the server(s) which where computing services are delivered over a network on
are located at the Edge of the network. However, it is very demand independently from device type and location [22].
important to determine the potential Edge locations. Edge MCC facilitate the offloading process in a distributed way
location can bring performance improvement and financial from a computer to a centralized server. MCC provides
benefits to the service providers and stakeholders. If the resources in terms of storage and computation to mobile
Edge location is too far from the centralized core and too users. This goal can be accomplished by bringing resources
close to the subscriber, the Edge computing may become closer to the proximity of end users [7]. Hence, Edge servers
expensive and complex. Edge location may vary according would improve the performance of MCC environments in
to application requirements. All the location-based contents terms of energy consumption, latency, and congestion. Some
should be hosted at some aggregation point. For example, of the motivations of MCC are given as follows [23]:
the services providers can choose an Edge location based 1) To prolong the network life time in term of energy
on the subscriber’s interest of videos. All the videos which consumption
are highly requested by subscribers should be cached at 2) Diverse application services

VOLUME 6, 2018 73469

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

FIGURE 4. Mobile cloud computing architecture.

3) Efficient utilization of resources

4) Virtualization
5) Latency
The major purpose of MCC is offloading tasks from a
mobile device to the cloud servers in order to overcome the
storage and computation limitation of mobile devices [8].
Moreover, the mobile devices also have the issue of battery
drainage. Therefore, the objective of MCC is to save energy
and prevent the factors that cause battery drainage [24].
An architecture of MCC is shown in Figure 4.

B. CLOUDLET
Cloudlet [25] has been introduced as an extension of MCC
and is considered as one of the keys enabling technologies for
MCC [26]. Cloudlet is a part of MCC that overcomes the high
WAN latencies. Offloading to the cloud is not always a solu-
tion, because of high WAN latencies. Real-time applications FIGURE 5. An example of Cloudlet [78].
need low latency and may not be achieved by offloading tasks
to the cloud. Tasks running on mobile devices may require
high computation and lower latency. However, the limitation Figure 6 depicts the Cloudlet reference architecture. The
of mobile device cannot fulfil these requirements. Therefore, main elements of the architecture are Mobile Client and
to provide computation power and to meet lower latency Cloudlet Host. A Discovery Service is a component run-
requirements, these tasks can be offloaded to the Cloudlet ning in the Cloudlet host and publishes Cloudlet meta-
instead of Cloud servers. Cloudlet is kind of small cloud data (IP address and port number). The mobile client uses
server located between mobile device and central cloud. IP address and port number for the specification of Cloudlet
Cloudlets are placed nearby to mobile devices with single- and to offload the computation. Once the Cloudlet is deter-
hop proximity and works as virtual-machine (VM). In the ini- mined for offloading, the mobile client sends the application
tialization phase, the framework requires the cloning (replica) code to the Cloudlet server. The Cloudlet server deploys the
of the mobile device application processing environment to a application code in the guest VM. Once deployment is done,
remote host. The entire application is offloaded using VM as the execution of the application is launched.
an offloading mechanism. The VM isolates the guest software Cloud servers are accessible via WAN technologies that
from the actual Cloudlet environment. The mobile device causes the latency issues for some applications. To overcome
serves just as user interface, whereas the actual applica- the latency issue in WANs, Cloudlet can be used by offloading
tion processing is performed on the Cloudlet infrastructure. mechanisms to nearby Cloudlet. Moreover, issues like energy
Figure 5 illustrates, that Cloudlets are distributed Internet can be addressed by Cloudlet [28]. Energy can be saved by
infrastructure components. Instead of accessing the distant offloading all the tasks to nearby resource rich Cloudlet rather
cloud servers, the nearby mobile devices can exploit the stor- than sending to the main cloud. Bandwidth is also a factor of
age and computation resources of Cloudlets. Such Cloudlets delay. Cloudlet could also help regarding bandwidth because
are deployed at public places such as coffee shops. wireless LAN bandwidth is typically two orders of magnitude

73470 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

performances are crucial on latency. This consortium is work-

ing on architectural design and testbed cases. There are no
restrictions for the contributors/researcher and everyone can
contribute irrespective of region and it is a public-private
ecosystem. Many use cases have been published such as con-
trolling the drone traffic [40]. Drone traffic can be generated
by IoT devices and an instant process may be required which
can be provided by Fog computing devices located near to end
users. The Fog technology are also commercially available
such as IOx and Local Grid [15].
Fog Computing addresses issues in the applications that
are distributed instead of centralized cloud architecture [41].
Due to proliferation of smart devices and their requirements,
cloud servers cannot be a single solution. Since all data
are gathered to one data center and required data are dis-
tributed to devices from the center. Therefore, the centralized
structure of CC causes a long delay by accessing WAN due
to congestion of data, which may create challenges in IoT
applications requiring real-time services. For example, Appli-
cations such as Augmented Reality (AR) services require
minimum latency for accessing computation resources [42].
In addition to the latency issue, many of the devises in IoT
are mobile, mobility management for the devices is hard to
be supported in conventional cloud architecture because of
FIGURE 6. A reference Cloudlet architecture [93]. centralized architecture. On the other hand, Fog computing
can be used for localized services according to the application
demands [43].
higher than the wireless Internet bandwidth available to a One comparative example of the localized service using
mobile user. Cloudlet is accessible at one hop. Therefore, Fog computing is illustrated in Figure 7(a) and Figure 7(b).
it can work even if the Internet is not connected for the access As illustrated in Figure 7(a), a mobile user wants to get
of cloud servers [29], [30]. In 2015, the Open Edge Com- flyers of store inside shopping center. However, to retrieve
puting (OEC) [31] has been proposed by telecom operators the flyers the mobile user needs to request the remote cloud
in collaborating with academia with the objective to leverage server. Since the store need to upload flyers on the cloud over
Cloudlets. For the better services of Cloudlet, the location and Internet. This results in high latency because of long physical
user identification of Cloudlet is very important [32]. No stan- distance between the Cloud server and the user. In order to
dardization effort has been made so far for the deployment of overcome such latency, the flyers should be stored locally
Cloudlets [33]. inside the shopping center so that user can access directly
instead of requesting to remote cloud server. As illustrated
C. FOG COMPUTING in Figure 7(b) the Fog server is deployed inside a shopping
CISCO proposed Fog computing model for the better man- center for providing localized services to users. From the
agement of the Clouds by enabling services, applications, and example in Figure 7(b), it might be obvious that Fog comput-
content storage in close vicinity to mobile end users [34]. ing provides better services than cloud computing in term of
In the Fog computing paradigm, data processing happens latency. However, some of the application cannot be satisfied
locally rather than being sent to the Cloud servers [35]. by Fog computing such as smart grid which requires global
Fog servers are located at base stations, streets, parklands, cloud servers. Therefore, the interaction among cloud and
restaurants, and shopping malls [1]. Fog computing sup- Fog servers should be regulated. Face identification [44] is a
ports offloading, caching, location awareness, and mobility use case that utilize both Fog and cloud servers and utilizing
information. It has many advantages for applications that benefits from both ends.
are delay sensitive [36]. According to CISCO [37], real- Figure 8 illustrates the use case of all the aforementioned
time requirements of IoT devices can be fulfilled by Fog services of face identification. When a user requests a service
computing paradigm because it provides services like com- to the Fog server using image as an input, then the subservices
putation, storage, and caching closer to the end user(s) than i-e face detection and image preprocess is executed by the
the conventional CC. Fog computing is a trending topic in Fog server. The pattern recognition is the last step and has
industries. The OpenFog Consortium [39] was formed in high computation complexity than the other initial services.
November 2015 with the objectives to solve the issues related Therefore, Fog server offload this task to the cloud server for
with IoT real-time applications and Tactile Internet whose further processing. When the cloud server processes the task

VOLUME 6, 2018 73471

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

FIGURE 9. Content forward process using MEC server.

for mobile networks only. This new paradigm of MEC

can accommodate and optimize the performance of many
applications.
MEC empowers the Edge of the network to perform in
isolated way from the rest of the network and allow access
to local resources [34]. The motivation behind MEC is the
proliferation of smart phones and the traffic generated from
the phones. According ETSI, MEC can reduce the latency
and can provide location awareness to mobile users. Require-
ments such as bandwidth (higher), latency (lower) and mobil-
ity should be met in future mobile networks such as 5G
and beyond. Therefore, to fulfil such requirements both the
RAN and core network should be optimized to serve billion
of devices [45]. It is noted that 5G system shall be able
to provide end to-end latency less than 10ms and 1ms for
some special cases [46]. MEC is the promising paradigm to
reduce latency. Furthermore, Edge servers address issue of
FIGURE 7. Localized services by Fog computing: (a) Retrieving flyers from congestion at the core network. The reason is that most of
Cloud, (b) Retrieving flyers from Fog.
the traffic is processed locally instead sending to the back-
bone networks [47]. Providing cloud-resources at the Edge
creates issues such as security and privacy [48]. Furthermore,
MEC servers could be congested due to loads on MEC
servers. Therefore, the non-functioning of MEC servers may
result in huge cost for operators [49]. In the following sub-
section, we discuss the content delivery process in the con-
ventional MEC architectures.

E. CONTENT DELIVERY IN MEC

Figure 9 depicts the data/service delivery process in MEC
architecture. The architecture consists of User Equipment
(UE), base station with MEC server installed, the core net-
work, and the Cloud. First, UE requests some data/service
FIGURE 8. Face identification using cloud server and Fog server. from the MEC server, which is installed at the base station.
The first time, the data/services are not found in the MEC
server. Therefore, the MEC server forwards the request to
and achieves the face identification, the result is then forward provider for requested data/service through the core network.
to the Fog server. The provider replies with data/service and sends back to the
MEC server. The MEC server saves the data as well the
D. MOBILE EDGE COMPUTING service subject to the policy of MEC. Now in the future,
The term Mobile Edge Computing (MEC) is coined by Euro- whenever UE requests the same data or service, then UE will
pean Telecommunications Standards Institute (ETSI) with the obtain it from the MEC server instead from the Cloud. This
aim to push computational power into RAN and to leverage reduces the latency and traffic for future requests. However,
the virtualization of software at the radio Edge. In order to without MEC installed at base-station, the user interacts with
realize the power of location both fixed and mobile networks cloud center via access point and core network. Therefore,
are accepting MEC. However, initially MEC were intended it creates an increased data traffic towards cloud and results in

73472 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

the operations of mobile network and offloading the sub-

scriber’s tasks. MEC provides access within RAN instead of
core WAN to minimize latency and to decrease the energy
consumption [51]. However, Cloudlets and Fog computing
provides the services to offload the subscriber’s tasks. Service
providers also affect the Edge server operations. Furthermore,
mobile network operators maintain the MEC and provide
services to a group of subscribers. However, Cloudlet or Fog
servers can be deployed with in a private environment such as
FIGURE 10. MEC progress. shopping mall and restaurant etc. [3]. Some of the key aspects
of each proposal are discusses as follows:

delay and high response time. Using Edge computing at base A. DEPLOYMENT LOCATION
station brings computation capability, and cache capability at The specifications of MEC states that MEC servers should
the Edge of network in close proximity to end users. be co-located with the cellular network base station. On the
other hand, Fog servers are generally provided by private
F. NO LONGER MOBILE EDGE COMPUTING environment such as shopkeeper etc. However, they can be
As previously discussed the term MEC was coined by ETSI deployed as routers and gateways in Internet service provider
in 2014 with a white paper authored by Huawei, IBM, Intel, (ISP) infrastructure. Cloudlet can be deployed in a distributed
Nokia, NTT DOCOMO, and Vodafone [50]. ETSI main way. Said otherwise, there is not exact location or vendor
objective is to shift storage, processing, and control to the for the deployment of Cloudlets. MEC server is reachable
Edge of network. While the term specifically referred to via third generation/Long-Term Evolution (3G/LTE) base
mobile networks initially, now ETSI’s scope covers both station. Therefore, MEC has the largest coverage area among
mobile and fixed networks. The MEC acronym no longer the Edge computing proposals. However, Fog servers and
refers to ‘‘Mobile Edge Computing’’ and instead stands Cloudlets are accessible via wireless access point (AP) whose
for ‘‘Multiple-access Edge Computing’’ [18]. The extension coverage area is much smaller than 3G/LTE. Mostly the
includes non-cellular technologies and now Wi-Fi is included Cloudlet study emphasis on Wi-Fi as an access technology.
within MEC’s scope. Most of the mobile operators uses Wi-Fi However, Cloudlet is not limited and can be applied in other
and most of the traffic accounts for Wi-Fi network on mobile wireless technologies [50]. Furthermore, Worldwide Interop-
devices. Hence, this is a very good inclusion for mobile erability for Microwave Access (WiMAX), is also supported
networks. by Cloudlet and may be utilized for communication between
Cloudlets [52]. Due to technological enhancement in cellu-
1) MEC PROGRESS SO FAR lar networks, the operators emphasis on MEC technology.
The ETSI MEC progress is illustrated in Figure 10. The In other words, future cellular networks are more commonly
basic specifications are released by MEC during the first referred to MEC instead of any other Edge Computing pro-
stage (2015-2017). During the second terms (2017-2018) posal. Hence, for cellular networks, MEC is the de-facto Edge
the agenda was to extend it to mobile networks and to computing technology. The reason is that Cloudlet and Fog
strengthen the collaboration with other Edge computing ini- computing have short range communication such as Blue-
tiatives. Along with the standardization and industry body dif- tooth and Wi-Fi and hence cannot reach to the level of MEC.
ferent business and deployment models should be explored,
such a private owners, enterprises, and providers. Enhance- B. DEVICES AND USERS
ments in QoE are very important. From 2018 onwards the Mostly studies related to Edge computing specifically Fog
ETSI planned to deploy MEC across different operators. computing addresses the use cases of IoT and vehicle-to-
However, in practical it shall take time before commercial vehicle (V2V) communication. For this reason, the users
release [18], [19]. and devices in Fog computing is expected higher than the
Cloudlet. However, Cloudlet covers the IoT devices but not
IV. DIFFERENCES BETWEEN EDGE COMPUTING V2V communication. The users of MEC is much smaller
PROPOSALS because MEC only focuses on subscribers and providers in
In the previous sections, we discussed in detail about Edge cellular environment.
computing proposals. Since the vital role of Edge computing
is to bring computational resources to the Edge of network. C. TRAFFIC PATTERNS
However, implementation specification differs at different Due to huge number of users served in IoT and the traffic
aspects for each proposal. In this section, we discuss the key generated from all the users and diverse devices affect the
differences among all Edge computing proposals in depth. traffic patterns at the Edge eventually. The traffic or data gen-
The fundamental objective of MEC is exploiting the capac- erated from Fog enabled servers and Cloudlet are continues
ity of conventional cloud at the Edge for accomplishing data generated from sensors while the traffic of MEC servers

VOLUME 6, 2018 73473

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

TABLE 2. Features comparison of conventional cloud systems and edge computing proposals.

are intendment and non-continuous. The reason is the pricing ETSI Industry Specification Groups (ISGs)and MEC Proof of
polices and targeted use cases of cellular networks [54]. Concept (PoC) Framework [55].
Edge computing use cases varies due to users, server
deployment and vendors. The main objective of Cloudlet is
D. DEPLOYMENT COST
to reduce the delay for real-time applications such as AR/VR.
Deployment cost is also one of the key factors that varies in
A set of use cases and scenarios has been introduced by ETSI
each proposal. MEC servers is located at the base stations
ISGs in the Group Specification (GS MEC-IEG 004) [49] to
and the cost of server deployment at the base stations are
showcase the role of MEC for the improvement of QoS.
expensive than Cloudlet. The cost of Fog servers is not higher
than MEC because of available devices such as a wireless
V. EDGE COMPUTING USE CASES
AP and a router etc. [43]. Therefore, the Fog server(s) result
This section aims at the protentional benefits of Edge
in a minimum deployment cost than other Edge computing
servers in real-life implementations. In following subsec-
proposals.
tions, we discuss the use cases given in the Edge computing
literature.
E. NUMBER OF SERVERS
MEC servers can be installed at the base stations and hence A. COGNITIVE ASSISTANCE
server density is limited to the base stations. While the Cloudlet is used as an enabler for the real time cognitive
Cloudlet can be installed at any public place such as coffee assistance applications that runs on wearable devices such
shop, shopping mall and restaurants etc. Cloudlets mostly as smart glasses [50]. The main objective of Cloudlet in
uses wireless local area network (WLAN) as an access tech- cognitive assistance is to accomplish task with low response
nology. Therefore, the density of Cloudlet is much higher than time. A Cloudlet framework has been proposed in [57] as a
other Edge computing proposals. Fog server(s) deployment practical implementation.
is average and cannot be installed everywhere like Cloudlet.
The comparison of conventional cloud systems and Edge B. BODY AREA NETWORKS (BANs)
computing has been shown in Table 2. The fundamental role of a BAN is to monitor the collected
In order to enable MEC in innovative use cases, several data with low latency [58]. A huge amount of data is gener-
organizations provide implementations in accordance with ated in BAN that need storage and computational resources.

73474 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

Moreover, the data which has been generated in BAN is life monitoring system in which the sensors data are transferred
critical and need to be monitored in real-time. The collected to the sink node(s). The data is processed and then transferred
data can be stored and analyzed with low response time to the Cloudlet. A huge amount of diverse data is generated by
using Cloudlet and Fog computing. A huge amount of data IoT devices that need to be aggregated. Cloudlet can be used
is uploaded by sensors in the case of electrocardiography for IoT environments since Cloudlet services can interact
(ECG). In case of instant processing of ECG data, the Fog with IoT devices [4].
servers can be used to process the collected data with low
response time. The data is collected by Fog servers and G. VIDEO ANALYSIS
various operations could be performed such as data filtering Low latency and jitter are required mostly in video steaming.
and data aggregation [59]. This can be achieved by enabling services at the Edge instead
of centralized servers to eliminate the WAN congestion. For
C. HOSTILE ENVIRONMENTS instance, if Fog server is deployed inside a bus or train then
The data generated by sensors in hostile environments such video streaming with low latency can be provided using
as war can be compromised [60] In order to detect adversaries Wi-Fi [43]. Fog servers are storage capable and have com-
in hostile environments, Fog servers can be used to analyze putational resources to analyze video steams [68] . The video
the sensed data of sensors with low latency [61]. processing tasks can be offloaded to the Cloudlet to minimize
delay and jitter to a certain threshold. Moreover, Cloudlet can
D. LANGUAGE PROCESSING also be used in stadiums to replay the video for audience.
Language processing applications required large resources In stadium thousands of users request the video at a time
and continues Internet connection for the better processing and hence resulting in congestion. Therefore, Cloudlet or Fog
of real-time processing [62]. Mobile devices are not good node can be used to offer video streaming services at the Edge
enough to provide all these resources because of limited and closer to users to mitigate the congestion.
capacity. Cloudlet can be used to lower the load on mobile
devices and to provide all the necessary resources. The H. AUGMENTED REALITY
Cloudlet eliminate the WAN for all Cloudlet enabled appli- The proliferation of smart phones and wearable devices
cations. An application named Android smartwatch has been such as smart glasses boosted the popularity of AR appli-
implemented which transfers the collected speech data to the cations [68]. These smart devices are not capable of
Fog servers to be processed [63]. After processing at Fog handling computational complexity. Cloud Computing is bet-
servers, the data is transferred to the cloud servers for further ter option for providing computational power and storage
analysis. capacity. However, AR applications are latency sensitive, and
traditional cloud systems cannot fulfil the required latency
E. SMART GRID NETWORKS requirements. Therefore, Fog computing can be used to fulfil
The Edge computing may be used for the smart grid networks. these requirements such as an increased throughput and lower
In case of Advanced Metering Infrastructures (AMI) the data latency.
may be categorized in different classes such as normal data Let us take an example scenario which takes inspiration
and real-time data [64]. Normal data is transferred periodi- from the pervasive AR scenarios mentioned at [69] and [70].
cally and do not need real-time interactions. In case of real- This example scenario explains the pervasive AR between
time situation such as alarming situation in a grid, the data Carlos and Sally and is discussed as follows.
need to be processed with minimum latency in order to pre- Carlos is informed by his voice assistant to meet Sally and
vent the fire hazards. Therefore, the Fog computing may be then he moves towards his self- driving car. His smart glasses
utilized for load balancing in such cases. Smart grid works mapped the destination map and time of arrival. Upon reach-
on multi-tier architecture and needs both Fog and Cloud ing the destined location, Carlos steps out from his car and
architecture [37]. Main Cloud servers provide services for entered into a busy market. His face identification application
storing the data for months and years and for analyzing large identified Sally in the crowd. Both Carlos and Sally met at a
amount of data. Local data can be processed by Fog nodes as specific location and after meeting their glasses make a visual
well as MEC server(s) [65]. route towards a nearest cafe. Right at this time, Carlos and
Sally both receives an alert message about major earthquake
F. WIRELESS SENSOR NETWORKS AND IoT to happen. After the alert message, the most structurally
Wireless sensors are resource constrained in terms of storage strong spots are highlighted to them within the building, and
and computation resources and unable to perform complex they run toward nearest spots. As the power goes out and
functions except sensing and relaying data [37]. Fog com- parts of the building collapse, Carlos and Sally get separated,
puting can overcome the storage and computation limitations trapped in separate parts of the building. Existing sensors
by offloading user’s tasks. Moreover, wireless networks have connect with their glasses and now visually displayed as
interoperability problems because of heterogenous environ- places where first responders can access video and audio
ments. This problem can also be solved by deploying hetero- feeds. Both Carlos and Sally wait trapped under the rubble
geneous Fog nodes [66]. CitySee [67] is an environmental for the first responders to arrive [70].

VOLUME 6, 2018 73475

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

In the above scenario, the services are offered with con- the network. Exiting TCP/IP-based solutions have some
tinuous context changing, from Carlos leaving his home, weaknesses when designed to support cloud services at the
to getting into the car, to meeting Sally, experiencing the Edge of network.
earthquake, and reconnecting with her. In this scenario, when Therefore, following are some of the limitations of
the earthquake occurs, pervasive AR is able to operate even TCP-IP based solutions for Edge computing and is discussed
the infrastructure is failed and the local resources has been as follows:
compromised. Therefore, the key requirement such applica-
tions is fast information response time which is unable to A. LATENCY
achieve with the existing host-oriented approach. Therefore, TCP-IP-based approach creates unacceptable latency
it is important to introduce emerging and innovative architec- for many latency-critical mobile applications such as
tural and networking paradigms to accommodate such latency autonomous driving, Realtime online gaming, VR, AR, Tac-
requirements. tile Internet etc. Such applications may require tactile speed
with latency approaching 1ms. Latency cannot be avoided,
I. CONNECTED VEHICLES AND SMART TRAFFIC LIGHTS
of course. However, it can be precisely measured, understood,
High mobility and unreliable wireless connection are the and managed in order to minimize its impact on QoE and
main reasons that affects vehicular communication [70]. The network performance [97].
performance can be improved by installing Fog technol-
ogy for communication between vehicles. Moreover, Fog
B. LACK OF BUILT-IN MECHANISMS TO DISCOVER
technology can also be used in smart cities such as smart
POTENTIAL PROVIDERS
lights to avoid the accidents by communication with each
other [4]. In addition, in road traffic signals scenario a video TCP-IP-based cloud platforms typically implement a Rep-
system can detect an ambulance and can change the traf- resentational State Transfer (REST) model to access cloud
fic lights for opening a way for ambulance. To avoid the resources through Hyper Text Transfer Protocol (HTTP).
road hazards and traffic congestion, an instant communi- HTTP uses Uniform Resource Identifiers (URIs) to identify
cation is needed with the driver of vehicle. Therefore, the the resources via Resource Discovery. When the resource
MEC technology can be used to provide roadside information discovery servers are not available then multicast DNS can be
to travelers by cooperating with sensors [55]. The sensors used for service discovery. The reason is that network layer
at the roadside senses the data and then transfer to MEC and transport layer in TCP/IP are unable to autonomously
servers for further processing. This collected data can be discover the resources defined by the application-layer names
pass on to other vehicles in the range. Using MEC technol- and do not allow application semantics into network layer
ogy the vehicle drivers can be informed of traffic conges- packets. This can be done by moving some of the functionali-
tion, accidents and other road hazards. Beside advantages ties implemented at the application layer (e.g., REST-related)
of MEC in vehicular communication, there are many secu- to the network layer [98].
rity issues such as reporting of fake data by vehicles, trust
and user privacy etc. [71], [72]. Therefore, security and C. DYNAMIC PLACEMENT OF APPLICATIONS AT EDGE
privacy should be considered for the protection of user’s NODES
data [73]. Dynamic placement of applications at Edge nodes and reso-
Even though the Edge computing offer better performances lution of requests to those nodes is a problem in EC. Further-
than cloud, however there are some limitations that needs more, the application software’s that run Edge computation
to be addressed. In the next section, we will discuss the must be first downloaded on the Edge-node, while mobile
limitations of Edge computing. clients request for resources from different locations. Mobil-
ity as well as diversity in user demand makes it very difficult
VI. LIMITATIONS OF EDGE COMPUTING to predict which functions would be requested in the future
The EC paradigm brings computational resources, stor- and from where in the network [99].
age and services from the cloud to the Edge of the net- In recent years ICN was proposed which addresses data
work and therefore closer to the consumers, in order to directly using content identifiers, instead of addressing the
reduce latency. Examples for such capabilities are resources host in the network. This fact allows mobility support
for computational-intensive and time-sensitive opera- by nature, while not maintaining network addresses of
tions or flexible deployment of applications and services. hosts. Moreover, it facilitates features such as in-network
However, the Edge Computing approaches (Specifically processing and caching of data. Therefore, to overcome
Multiple access Edge Computing) is based on virtual the above-mentioned issues, ICN in Edge computing is
machines (VM) and totally based on the host-centric net- promising.
working model (TCP/IP approach). This creates challenges We have discussed Edge computing in depth so far. In the
in data dissemination between the highly mobile users next section, we will discuss the futuristic paradigm called
as well as the addressing issues of domain name sys- ICN that would be used as deployment strategy for Edge
tems (DNS) due to nodes constantly joining and leaving computing.

73476 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

VII. INFORMATION-CENTRIC NETWORKING

Various ICN architectures have been proposed that share
common ideas and principles, such as name-based content
retrieval and discovery, content-based security, in-network
caching, and connectionless receiver-driven communication
models [74]. Interested readers are referred to [75] for more
detailed information about such projects and architectures.
In the ICN paradigm, a unique name is assigned to con-
tent, and that content is retrieved without knowing about the
location where it resides unlike traditional IP systems. That
name uniquely identifies the content (e.g., a video, a picture,
a document, a web page). In this paradigm, the content is
secured instead of communication channel/pipe. Connectiv-
ity between consumers and producers is not necessary for
the exchange of content. Anycast data retrieval is supported,
meaning that the router forwards requests to any node holding
content. ICN offers flat or hierarchal naming. The former
is easier to manage and self-certifying. However, it is not
readable to humans. The latter is human-readable and the
FIGURE 11. Simplified interest packet processing in ICN.
most usable and backward compatible [76]. Both have pros
and cons. If flat names are used, attackers can be avoided
because the names are un-readable to humans. If hierarchal However, due to rise of IoT and real-time applications such as
names are used, attackers know to attack sensitive data since AR, VR and Tactile Internet, it becomes challenging in terms
the information is unveiled to users. However, content-based of mobility, scalability, security and network management.
security is implemented at the packet level rather than at the ICN is a promising paradigm that is based on named based
communication channel/pipe level. communication rather than host based. The first research
ICN is based on two packet types, Interest and Data. The efforts in ICN was only limited to naming the content and to
communication process is as follows: request on the network layer. However, in today’s era the ICN
1) Consumers send Interest packet(s) containing the also provides the naming of services with the help of Named
names of the requested content. Function Networking (NFN) [100]. Therefore, the content
2) Data packet(s) flow back, carrying the named and and services can be named and requested at the network
secured content chunks, by following the same path layer without relying on the fixed communication of IP based
through which the Interest packets were sent [74]. approach.
Each node maintains three types of data structures: (1) a Moreover, Edge computing is able to deploys services on
Content Store (CS), which is capable of caching data tem- the edge of network. The edge computing provides local
porarily; (2) a Pending Interest Table (PIT), which retains the computing and storage, thereby reducing latency. In addition,
records of unsatisfied Interest packets (3) a Forwarding Infor- ICN natively supports decentralized caching, self-
mation Base (FIB), which traverses Interest packets toward authentication and multicast that can enable Edge computing
the data providers. deployment. Both ICN and Edge computing has some correl-
When a consumer node wants to access specific content, ative properties, such as decentralization, local storage etc.
it sends an Interest packet containing the name of the con- Fortunately, edge computing is able to support storage and
tent. When some relay node(s) receives the Interest packet, computing naturally. A combination of both ICN and edge
it first checks its own CS for data availability; if match is computing could result in better performance gains if paired
found, the node sends the data back to the consumer via the the features of both together.
same interface from which the interest is received. Otherwise, There are many expected benefits resulting from the inte-
it checks its PIT. If the entry is found in PIT, the node updates gration of ICN in Edge computing which are described as
the existing PIT table by adding a new incoming interface follows:
entry and discards the Interest for further processing. Other-
wise, a new PIT entry is created, and the Interest packet is 1) VIRTUALIZATION OF SERVICES
sent further via interface(s) stored in the FIB. The Simplified In the literature, support is available for virtualization of
procedure when a node receives an Interest packet is shown services, such as Docker [101], Amazon Lambda [102]
in Figure 11. or serverless computing technologies such as uniker-
nels [103]. These technologies are useful for Edge
A. WHY ICN IN EDGE COMPUTING? computing. Since it provides encapsulation of functions into
In the conventional Internet design the communication hap- self-contained software components, executable on Edge
pens between fixed entities due to host based approach. nodes and totally independent of its deployment structure.

VOLUME 6, 2018 73477

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

Moreover, virtualization technologies are lightweight and 7) IN-NETWORK SECURITY

efficient, leveraging their benefits to provide improved Edge Protection and trust are implemented at the packet level,
services in resource constrained environment. rather than at the channel level. By design, ICN, offers native
support for security, which are still not effectively available
2) FUNCTION NAMING in the host-centric paradigm [104].
In NFN [100] ICN function naming has been proposed The self-certify names model of ICN enable to verify the
to identify network resources and to support the execution binding between public key and self-certify name in dis-
of network functions. Explicitly named functions can be tributed system without relying on a third party. This can
resolved in network nodes, while network-layer requests reduce the security risk of involving a third party. However,
(i.e., Interests) can carry input information for Edge- it is difficult to maintain the centralized key management
executable functions. Function code can be stored in node infrastructures such as Central Authority (CA), especially
caches and migrate across the network following user in the constrained IoT. The reason is large communication
demand. and computational overheads incurred due to complex trust
These solutions can be enhanced to support Edge com- chain of certificate verifications. The problem was solved by
puting by making use of ICN technologies to allow devices Zhong et al., with a distributed key management scheme
to express the services they need without having to specify adopting identity-based public key cryptography (IB-PKC)
the exact node that could provide those services. They could [119] to avoid the problems of single location. However,
provide support by: a) route the requests to the best Edge it suffers the key-escrow problem and single point of failure.
server/device; b) instantiate/migrate the network function to To solve the key-escrow problem and single point of fail-
Edge device that is closer to the user. ure, Al-Riyami et al., introduced a certificate-less public key
cryptography (CLPKC) [120]. In this scheme, the private key
3) INTEREST AGGREGATION is eventually generated by users and key generation center
In ICN over Edge computing, the interests can be aggre- together, and the attacker is not able to acquire the private
gated with the same name coming from difference con- key of any users even when Key generation center has been
sumers, so that they cannot be duplicated over a given link compromised.
towards a producer or service executor. However, the legacy Recently Jun Wu et al., propose an anonymous distributed
IP approach does not allow aggregation of requests [104]. key management scheme based on CL-PKC specifically for
Space Information Network (SIN) in order to overcome the
4) CACHING OF RESULT/COMPUTATION FOR REUSE security issues [121]. Authors designed a distributed key
Nodes in NFN caches the results of functions/services and management system model for key exchange services. Since
make them available to other consumers without performing authors scheme is based on the certificateless public key cryp-
the computation again and again [100]. Moreover, the dis- tosystem, therefore, it can avoid the problems of complicated
tance between the Cloud and Edge networks can be several certificate management and key escrow. Furthermore, imple-
kilometers [87], which may result in a significant delay. mentation methods have been provided for the generating
ICN over Edge networks can achieve latency requirements and updating of key pairs. The security analysis and com-
by providing data and services that are close to end users via parison of computational overhead confirms it’s security and
caching the content and results as well. less computing cost as well. However, this was specifically
designed for SIN, and may be used for NDN based Edge
5) LOCATION-INDEPENDENT NAMING computing.
Hierarchical user-friendly URI-like names uniquely identify Liu et al. [122] proposed a scheme for information-
a content (e.g., a movie, a picture, a song) as well as context centric social networks (ICSN) and claimed that the existing
(location, identity etc), independently of the identity/locator schemes for the conventional social networks cannot fulfill
(i.e., the IP address) of the node generating/hosting it. There- the requirements of ICSN. Therefore, a fog computing-based
fore, ICN is not bound to specific address of content, ser- content-aware filtering method for security services, FCSS, is
vice or context. In ICN those all reflect the named pieces proposed in information centric social networks. They intro-
of content and could be requested via name directly at the duced fog computing in IC-SN, and the content aware fil-
network layer. tering scheme is proposed for security services. Such Edge
computing based ICN solutions can be introduced in many
6) NATIVE IN-NETWORK CACHING NDN-based Edge computing applications which are detailed
By integrating data caching into the forwarding process, in Section VIII. Moreover, CL-PKC is one of the areas to be
NDN makes content delivery more robust against packet explored for NDN based Edge computing applications.
losses and improve content availability. Besides content
caching it is also useful for function results that are executed 8) BANDWIDTH UTILIZATION
in the network. Therefore, ICN not only provides content Storage and bandwidth can be efficiently utilized due to ICN
caching but also functions/code caching in order to avoid re- multi-point delivery mechanism. The content will be sent to
request the content or re-execute the function/code [69]. a group of users and will not be unicasted. With growth of

73478 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

users the bandwidth will not be affected too much unlike difficult to express only through simple expressions and
unicasting. Therefore, using ICN over Edge computing we acquiring additional function code.
could utilize the bandwidth affectively. In [123], Named Function as a Service (NFaaS) has
been proposed that supports more sophisticated processing
9) BUILT-IN MOBILITY SUPPORT with lightweight virtual machines in the form of named
Consumer Mobility is a built-in feature of ICN due to receiver unikernels. The unikernels are actually the functions/codes.
driven and connectionless data communication nature. When In NDN the content is cached in the Content Store. However,
an end device moves to a new location, it simply needs to in NFaaS an additional data structure call Kernel store has
re-express request for its interested data. However, support been introduced. Every node contains Kernel Store that stores
for producer mobility is still a research problem in ICN. the unikernels. The Kernel store is responsible not only for
Some networks are highly mobile, such as vehicular net- storing functions, but also for making decisions on which
works (VNs). Therefore, if mobile users can only receive functions to run locally. Since the Kernel Store has lots of
the content from the provider (original server), then the functions and which functions to download locally to the node
connection can be lost during the mobility of users. Due is calculated by score function. The score function score all
to interruptions in connectivity, users will re-request the the popular function that is requested more frequently and
content from original server. In ICN the mobility feature the main goal of score function is to identify the uniker-
is inherently supported. The devices can directly commu- nels/functions that are worth downloading locally into the
nicate using service names instead of specific host such as node’s memory.
Netflix.com or Youtube.com. Services are provided by the Amade et al. [2] extended the NDN architecture to turn the
network and does not rely on end to end communication. network edge into a dynamic computing in the IoT domain.
ICN caching provides a copy of the content to all users and The proposed scheme performs distributed in-network
hence mobile users no longer make requests of the original IoT data processing at the network edge, by relying on NDN
server. Therefore, if the mobile users keep moving in network augmentation and named computations. This scheme also
as in VN, then content can be obtained from the nearest cache performs the dynamic execution of services, according to the
instead of going to the original server. Hence, a reduction interests popularity function. In the proposed scheme authors
in delay and support for mobility is achieved. Moreover, have performed minor modification of legacy NDN and used
ICN offers heterogeneous wireless support where mobile a naming scheme that identifies IoT contents and services
devices should be able to transparently use a variety of com- without affecting the NDN routing.
munication technologies [105]. Amadeo et al. [125] proposed an NDN based scheme call
NDNe (NDN at the edge) that supports cloudification at the
edge. In NDNe the existing NDN packet is extended and
B. FEASIBILITY OF ICN DEPLOYMENT ON EDGE
names are used to address, not only ‘‘contents’’, but also
The traditional Internet is based on TCP/IP (host oriented) to identify different types of cloud service (e.g., storage,
network model. Therefore, to replace the TCP/IP model with computation, etc.)
ICN is impractical. However, there is a way to deploy the
ICN partially or overlay such as ICN over IP or IP over ICN. VIII. APPLICATIONS OF ICN INTEGRATION IN EDGE
Deploying ICN on edge service not only can help to miti- COMPUTING
gate the ICN whole-network deployment complexity, but also Future networks such as 5G will result in data rates of multi
makes the network model more flexible. The combination of gigabits per second and will support the scalability of devices.
ICN and EC is able to offer a great combo and such approach 5G systems will support real-time networks to deliver real-
may result in maximum performance. time controls. Latency is the fundamental unit of 5G systems,
and thus the target is to enable low-latency applications such
C. RELATED WORKS ABOUT ICN OVER EDGE COMPUTING as Tactile Internet, autonomous driving, industrial robotics,
There has been no focused attempt to adjust existing pro- and VR/AR applications. All these requirements cannot
posals for Information-Centric Networks to support edge be fulfilled by the existing TCP/IP approach due to host-
computing with the exception of few related works which are oriented approach and fall short for such applications. Indeed,
discussed as follows. ICN principles can directly address the above challenges.
Sifalakis et al. [100] proposed a pioneering scheme call In ICN network, the content is cached along the path. There-
‘‘Named Function Networking’’ (NFN) for the extension of fore, the requested content can be from the source node or the
NDN to the edge computing. In NFN, the name field of inter- other content caching nodes thereby reducing latency. Instead
est packet carry the name of the content as well as expressions of binding security to host node, ICN advocates the security
for named functions. The network is in-charge of computing model for the content. This model focuses more on securing
the result and resolving the forwarding plane of NDN. How- the content not the channel. Moreover, consumer mobility is
ever, NFN is constrained by the number of services/functions natively supported in ICN architecture. In 5G application it
it can support. In many scenarios, nodes require more is possible to have frequent handover events. Since ICN can
sophisticated processing, custom code and libraries, which is natively support mobility through its content-centric design

VOLUME 6, 2018 73479

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

and stateful forwarding plane, Therefore, an ICN solution in AR, dynamic content is visualized compared with today’s
for consumer mobility would be more appropriate. In addi- static augmentation [87].
tion, 5G networks are expected to utilize multiple interfaces. The existing AR applications are built upon TCP/IP
ICN offers heterogeneous wireless support where mobile protocol stack and rely on cloud computation. To enable
devices should be able to transparently use a variety of com- pervasive AR applications, it is important to explore new
munication technologies simultaneously. computing paradigms, new approaches such as ICN to net-
Following subsections details some of the 5G applications work communications. Though the existing AR applications
as follows: bring utility in specific domains, their reliance on cloud ser-
vice may limit the potentials of AR [89]. Therefore, Edge
A. TACTILE INTERNET computing paradigms, within physical vicinity, could achieve
Low latency and the highest reliability of data with more the required low latency while protecting user privacy. Edge
security for real-time systems, such as real-time gaming, computing paradigms are important in accomplishing perva-
industrial automation etc., will be provided by Tactile Inter- sive AR. To support Edge computing, ICN, can be introduced
net [12]. 5G wireless solutions will fulfil the requirements that how ICN could address the requirements of resource
of wireless communication for 2020 and beyond. There- discovery, multicast support for context-content exchange,
fore, 5G is predictable to support the Tactile Internet at and experimentation with user experience.
the Edge of wireless networks. To reduce the latency and These applications will drive the ICN and Edge computing
bandwidth requirements, content will need to be either local- design in terms of latency, performance, and scalability and
ized or pushed to the Edge of networks [86]. would prove the importance of ICN in Edge computing. The
ICN naming mechanism may result in lower signaling costs
B. INDUSTRIAL AUTOMATION in content retrieval [89].
Various control processes exist and require different data This article intends to show ICN as an important deploy-
rates, reliability, security, and latency [87]. Automation ment strategy for the utilization of ICN mechanisms and
industry applications will be driven by 5G systems. Cur- their key benefits in such environments of Edge computing.
rently, the wired industrial Ethernet is used to control pro- ICN can be beneficial, allowing new concepts to be devel-
cesses. Wireless systems adoption is necessary for flexible oped based on named requests, caching, any-casting, and new
production, which requires guaranteed reliability and low applications to be created, as a true future Internet paradigm.
latency [88].
IX. STANDARDIZATION ACTIVITIES
C. AUTONOMOUS DRIVING Many academic, industry and standard organizations are
Within the context of 5G, autonomous driving is discussed working to address various research topics in Edge net-
as a new phase in mobility. Today’s applications require works and ICN. In the United States, these programs come
latency to be less than 10ms for vehicle safety to avoid under the National Science Foundation (NSF) Future Internet
collisions. Therefore, if the bi-directional data exchange for Architecture initiative (e.g., CCN, NDN, Mobility First, Xia).
the movement in autonomous driving is considered, a latency In Europe, they are under the European Union (EU) Frame-
of millisecond will likely be desired [87]. Highly reliable work programs (through H2020).
and proactive behavior is thus needed in future 5G com- ETSI plays a vital role in the development and imple-
munication systems. However, ICN with Edge computing mentation of telecommunication standards. ETSI is currently
could solve the issue of latency and mobility for autonomous working on the development of standards relating to cloud.
driving. The work of ETSI MEC aims to provide IT and CC capa-
bilities within the RAN. MEC shall enable applications and
D. INDUSTRIAL ROBOTICS services to be hosted on the base station. These applications
Autonomous robotics may react in an irregular manner if and services can benefit from being in close proximity to the
real-time communication fails. Therefore, it may lead to an customer and from receiving local radio-network contextual
unwanted behavior. There are many scenarios in the manu- information [2].
facturing of robotics that necessitate a maximum delay target The ETSI MEC ISG was founded in December 2014 and
of 100µ s and round-trip reaction time of 1ms [88]. In order to the purpose of the ISG is to create a standardized, open
provide such latency, it is of upmost importance to bring the environment which will allow the efficient and seamless inte-
content and resources closer to the user. Moreover, integration gration of applications from vendors, service providers, and
of ICN with Edge computing is a promising approach for third-parties across multi-vendor MEC platforms.
providing real-time communication. MEC ISG has been published the following specifications
so far [56]:
E. VIRTUAL AND AUGMENTED REALITY 1) Foundation specification GS MEC 001 MEC Terminol-
Tactile Internet can prove helpful for VR and AR appli- ogy (2016-03)
cations. Many users mutually perform tasks by perceiving 2) Foundation specification GS MEC 002 MEC; Techni-
objects using simulation tools in VR. On the other hand, cal Requirements (2016-03)

73480 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

3) Foundation specification GS MEC 003 MEC; Frame- Next Generation Mobile Networks (NGMN) published a
work and Reference Architecture (2016-03) 5G white paper [85] in March 2015. The paper provides
4) GS MEC-IEG 004 MEC; Service Scenarios (2015-11) requirements for 5G and encourages the adoption of new
5) GS MEC-IEG 005 MEC; Proof of Concept Framework emerging technologies. ICN is one of the technology building
(2015-08) blocks considered by NGMN. ICN is described as having
In [79] and [80], Edge computing is demonstrated using the potential to migrate from a host-centric and node centric
OpenStack to bring the cloud closer to the mobile Edge. The model to a content-centric, data-oriented and information-
Open Edge Computing project [31] was created in 2015 to centric model with an intrinsic focus on named information
facilitate prototyping applications that can take advantage objects in network caching and name-based routing.
of Edge computing and engaging with relevant development
communities. X. VENDOR SOLUTIONS
Open Fog Consortium (OpenFog) is founded in In this section, we will discuss about vendors which are
November 2015 by a major industry movers and leading working on the Edge computing solutions. This subsection
academic institutions. The OpenFog Consortium [40] was explains the up to date hardware and software solution related
founded by the following technology industry leaders: ARM, to the Edge computing paradigm so far at the time of writing
Cisco, Dell, Intel, Microsoft and Princeton University in this survey paper.
order to solve the bandwidth, latency and communications
challenges associated with the IoT, Artificial Intelligence,
A. ADLINK TECHNOLOGY
Robotics, the Tactile Internet and other advanced concepts.
3rd Generation Partnership Project (3GPP) [80] recently ADLINK technology [106] is working towards hardware
started the NextGen study item, which aims at defining solutions for Edge devices which are capable of Fog comput-
the 5G system architecture. ICN is not yet being explicitly ing and MEC features. Recently a product has been launched
addressed as a dedicated payload type in the 3GPP system. by ADLINK called SETO-1000 [96] which is a part of the
This is partly due to the fact that the related work on defining MEC architecture. SET0-1000 is designed with the aim to
the ICN protocol and the related mechanisms in Internet provide networking solutions in extreme outdoor environ-
Research Task Force (IRTF) and Internet Engineering Task ment. It is providing CC features within the RAN closer
Force (IETF) are still ongoing. to mobile users. The platform consisting of storage capac-
The key enabler for many ICN use cases is the ability to ity, accessibility to real-time radio and network information.
deploy ICN routers close to the radio network. Information- Moreover, it reduces the backhaul cost and increase network
Centric Research Group (ICNRG) was formed to identify efficiency in terms of delay and bandwidth since the data is
outstanding research challenges for ICN, and to couple processed at the Edge.
ongoing ICN research with solutions that are relevant and ADLINK is a Premier Member of the Intel Inter-
appropriate for evolving the Internet at large [81]. The net of Things Solutions Alliance. Moreover, ADLINK is
work in the ICNRG (in the form of Informational Request contributing in many standardization projects, including
for comments (RFCs), meeting contributions, etc.) is com- Open Compute Project (OCP), ETSI MEC , Network function
pletely documented and accessible through its website at virtualization (NFV), OpenFog Consortium, Telecom Infra
https://trac.tools.ietf.org/group/irtf/trac/wiki/icnrg. Project (TIP), the PCI Industrial Computer Manufacturers
ITU Telecommunication Standardization Sector (ITU-T) Group (PICMG), the PXI Systems Alliance (PXISA) and the
created a group IMT-2020 [82] to study how emerging 5G Standardization Group for Embedded Technologies (SGET).
technologies will interact in future networks. This group
also included studies on high-level network architecture. The B. ADVANTECH
Alliance for Telecommunications Industry Solutions (ATIS) Advantech [107] enables MEC and OpenFog deployments by
board initiated a committee to investigate the Evolution to providing hardware solutions at the outer Edge, and pushes
Content Optimized Networks (eCON) [83]. Although not the processing, storage to the far Edge. Recently, Packetarium
exclusively focused on ICN, most work addressed the overall XLc, which is a virtualized platform has been introduced
ICN opportunity space from a network operator’s perspective. by Advantech for Edge-computing deployments. Moreover,
5G Americas present a 5G white paper [84] which describes it offers solution for 5G networks based on open architec-
some detail ICN architecture, along with ICN benefits and tures and industry standards. According to [107] Packetarium
use cases. ICN is presented as a potential technology for XLc is installed far from centralized data centers, and it
consideration as 5G. The paper suggests that 5G should be can support 9 slots, up to 288 Intel Xeon processor cores.
based on new network architectures and protocols designed Advantech along with Vasona, Brocade, GigaSpaces and
specifically, with support for mobility, security and content Saguna Networks takes part in an ETSI MEC PoC, ‘‘Multi-
caching as fundamental design criteria. ICN as realized in the Service MEC Platform for Advanced Service Delivery’’. The
NDN and Content-Centric Networking (CCNx) programs is PoC illustrates how the infrastructure between a Network
described as a leading architecture that can meet such design Function Virtualization (NFV) and a cloud system can simul-
criteria. taneously support many MEC platforms and applications.

VOLUME 6, 2018 73481

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

The MEC platform offers operators optimization of RAN the rate and distribution of traffic (over time). The great
performance, particularly, to minimize latency. advantage of open caching, data which is frequently in use,
can be accessed without requesting any action from CDNs,
C. ARTESYN content providers or subscribers, since it is stored at the Edge
Artesyn [108] focuses on virtualized solutions in the RAN of network.
as well as in the core network. Artesyn is also designing
the hardware for next generation networks to facilitate the F. VASONA NETWORKS
transition to 5G. The MaxCore platform has been developed Vasona Networks [111] is working towards Edge computing
for Edge computing. Artesyn, with the support of MaxCore to manage the traffic at the Edge between the mobile core
platform, offers a power efficient, scalable and flexible fully and the RAN, to optimizing RAN performance and QoE. The
integrated suite of cloud-based products. main aim of this technique is to assist mobile operators by
MaxCore is mainly focusing on optimizing performance in providing them better end-user quality and efficient use of
terms of latency and bandwidth, for the environments where network resources.
the subscribers and traffic is in high density. Artesyn has Initial solutions by Vasona, focused on the challenges of
worked on many use cases such as Known-location services video traffic and the requirements of its latency, that mobile
(for government, retail, education and health), IoT applica- operators have to suffer. Standards-based software platforms
tions (for smart cities) and AR. has been developed by Vasona for MEC that could be placed
at an aggregation point between the mobile core and the RAN.
D. INTERDIGITAL Typically, a thousand or more cells can be covered by locating
In order to enable ultra-low latency, real-time and the MEC functionality in an aggregation point.
location-specific traffic optimization and context awareness, Today, Vasona has two products (i.e. Smart AIR and
Interdigital [109] is working towards MEC, to bring storage Smart VISION):
and Processing at the far Edge of user devices. Interdigital 1) Smart AIR is referred as an Edge application controller
actively takes part not only in the ETSI MEC i-e the Open developed at the individual cell level. In the case when
Fog Consortium, open-edge initiatives, groups (for exam- RAN is overflowed, SmartAIR works to control the
ple 3GPP), but also in advanced wireless research (PAWR) individual traffic flows at real time to overcome latency
platforms such as NSF’s Platform and European Commis- and offers efficient network utilization.
sion’s Horizon 2020. Furthermore, in collaboration with Uni- 2) SmartVISION is described as to assist operators to
versity of Essex and Intracom, Interdigital also gets involved analyze RAN performance, by providing them real-
in an ETSI MEC PoC. Interdigital integrates ICN and SDN. time and historical data. SmartVISION has an ability to
According to Interdigital, the paradigm shift of networking obtain information on user activity, content usage and
away from host-to-host communications, to content- and application, for every cell sector. The optimization and
name-based addressing by ICN, is mandatory to satisfy the planning for network performance and expansion can
requirements of 5G’s latency. Furthermore, Interdigital keeps be done based on this information.
focusing on its development and research, to combine the
services and network infrastructure, such as converge of NFV, XI. FUTURE RESEARCH DIRECTIONS FOR
MEC and SDN with the evolution of 4G and the development ICN OVER EDGE COMPUTING
of 5G. In this section our aim is to direct the researchers for the
enhancement of ICN and Edge computing. Although ICN
E. QWILT paradigm promises to provide the required features currently
Qwilt [110] was established in 2010, to facilitate the broad- not addressed by the existing 5G research. Therefore, the fol-
band fixed and wireless services. Qwilt offers optimization lowing question arises: How can ICN be enabled in Edge
both in terms of latency requirements and capacity of high computing? How can ICN be combined and co-optimized
video traffic loads to enhance the QoE. with these networks? Several few issues are described as
In order to optimize the video delivery and to facilitate follows:
real-time applications such as VR and AR, Open Edge Cloud
platform offers content delivery solutions and open caching to A. MORPHING (IN-NETWORK DATA MANIPULATION)
service providers at the network’s Edge. To minimize latency, The Internet of Things (IoT) promises to connect billions of
the Edge Cloud platform enables the storage capabilities and objects to the Internet. It needs to support 50-100 Billion net-
computation as much as possible to the Edge of the network. worked objects, many of which are mobile. A large amount of
The main purpose of Qwilt’s Edge Cloud solution is to data will be generated by things and many applications will be
extend the Content Delivery Networks (CDNs) and content deployed at the Edge to consume these data. IoT is compatible
providers to reduce the costs of transport for efficient deliv- with Edge networks because of its IP nature. However, ICN is
ery. To use the network infrastructure more efficiently, Open not compatible with Edge networks, and hence it is a very
Caching software integrates open caching with analytics and challenging task to handle IoT devices and data at the Edge
media delivery, which allows them efficiently to manage of network using ICN. There are many potential challenges

73482 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

cific locations in the network to ensure a trade-off between

effectiveness and computational resource demands.

B. CACHING
Caching is very beneficial to increase the availability of
data and to speed up data retrieval. The cache everything
everywhere approach is not useful, resulting in the wastage
of cache resources and creating a considerable amount of
redundant data in the network. Data replacement policies that
will behave according to the behavior of the content and
interest are needed [92].
Caching creates the following three main questions:

1) WHAT TO CACHE?
It is useless to cache all the content in the network. Therefore,
it is necessary to consider the popularity of content and
determine what content to cache. Many users request the
same content, and therefore unpopular content has a negative
FIGURE 12. Morphing in ICN enabled networks. impact on the utilization of caching.

due to ICN integration in Edge networks, as discussed above. 2) HOW TO CACHE?

However, on the other hand, ICN in Edge networks can be It is necessary to evaluate the reputation of content rather
beneficial for IoT to support seamless mobility, security, and than applying traditional caching policies, such as the least
efficient content and service delivery. recently used (LRU), least frequently used (LFU), and first
As discussed in [91], applications in IoT will be deployed in first out (FIFO).
at the Edge, and there are no compatibility issues. By com-
patibility, we mean that the Edge networks and IoT are both 3) WHERE TO CACHE?
IP-based (host-centric) unlike ICN. Therefore, Edge net- Integrating ICN in Edge computing creates another challenge
works and IoT can co-exist. To handle a huge amount of raw for the deployment of caching. In current cellular networks,
data in the network of IoT, morphing is a strong concept. caches are deployed at two places, evolved packet core (EPC)
These raw data create congestion and delay in the network, and at the radio access network (RAN). EPC consist of packet
resulting in various issues. Before going into detail of these data network gateway (P-GW), serving gateway (S-GW) and
issues, we will first explain what morphing is? Morphing mobility management entity (MME) in Long Term Evolution
is a concept used for in-network data manipulation/filtering. (LTE). In 3G networks, RAN consists of NodeBs and evolved
Data inside the network are filtered by some intermediate NodeBs (NodeBs) in 4G LTE networks [93]. It is unclear
nodes and then sent to the sink node. In this approach, only whether to store in a MEC server or ICN node or both.
filtered data that are meaningful are sent to the sink node Moreover, to execute functions there is a strong need of
instead of sending all the raw data. This filtering of raw data availability. Since ICN provides in-network caching
data results in low power consumption, low traffic, and fewer capabilities, therefore, proactive caching strategies needs to
transmissions, improving the bandwidth and lifetime of the be investigated for prefetching data as well as functions to
network. Hence, only manipulated data are sent to the sink. speed up computation of service/function results on the Edge
A consumer is only interested in processed data rather than devices.
raw data. Therefore, we are directing interested researchers to
perform morphing in ICN-enabled Edge networks. However, C. DEPLOYMENT STRATEGIES AND ORCHESTRATION
there are some pros and cons that may result due to ICN Cloud technologies such as unikernels simplify the deploy-
integration. First, the ICN node does not provide any data ment of functions and services and therefore provide efficient
transformation (e.g., filtering, aggregation) [91]. However, data processing and dissemination [112]. The loosely coupled
ICN could enable lightweight in-network data manipulation addressing concept of ICN simplifies both the access to and
at intermediate nodes by embedding semantics awareness at the placement of functions and services in the network. This is
the network layer, as shown in Figure 13. due to the fact that addresses of physical components need not
Second, issues may result such as increased complexity be maintained and well-known by consumers as it is required
and function overloading inside the network if transformation in today’s host-centric networks.
is performed. There is space to think about morphing in However, resource allocation and management of func-
ICN-enabled Edge networks. Therefore, the requester can tions and services defines a new research problem. There is
retrieve aggregated data from the best node in the network. a need of distribution strategies for functions and resources
Note that data morphing should be carefully applied at spe- are unnecessarily occupied and thus decreases network

VOLUME 6, 2018 73483

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

efficiency. That means that most of the resources in the net- context-sensitive service results [69]. Furthermore, mecha-
work are underutilized. For efficient data dissemination such nisms such as name-based routing and forwarding should
information needs to be transferred from one service instance cope with the mobility of network participants.
to another according to the mobility model of the mobile
node. Such mechanisms need to be supported by QoS mech- F. DISCOVERY AND DELIVERY
anisms, required to differentiate between different types of Name-based routing (NBR) or look-up-based resolution sys-
data and services and to ensure efficient data dissemination. tems (LRSs) are used for content discovery and delivery in
Additionally, such strategies/policies must ensure fairness ICN. With NBR, the interest packet is sent in a manner of
among the different consumers and providers of the services. hop-by-hop transmission by forwarding node(s) by looking
up a name match into their FIB. Once the content is found,
D. SECURITY it follows the reverse path back. With LRS, the interest packet
In ICNs, security features are directly introduced as part of the is sent to a resolution system, which may vary subject to
content itself instead of the transport layer as given in today’s the ICN architecture. Different architectures may have dif-
IP networks. Enabling ICN in Edge networks creates many ferent resolution systems depending on the implementation
issues for security and privacy. ICN over Edge will share data of architecture [21]. Therefore, how data could be discovered
with all users because the ICN node has the capability of any and delivered in paradigm of ICN over the Edge computing?
casting. Due to the anycasting feature of ICN, data will be Hence, a new resolution system may be designed to discover
sent to all future interested users [91]. and deliver data and resources in ICN over the Edge comput-
Moreover, due to network caching, security is increased ing approach.
by the fact that data is expected to stay within untrusted The resource discovery could be of two types 1) Services
caching nodes [69]. This also includes privacy concerns while discovery: (that means which service(s) are running on the
requesting customized service results. In recent years, mech- Edge node and what are the capabilities of each Edge node.
anisms are proposed that are showing security features. How- 2) Resource availability: (that means, what is the current load
ever, such mechanisms are not addressing the requirements of on the Edge nodes and which Edge node is the best node
mobile scenarios. The question is how to exchange encryption for task execution in terms of central processing unit (CPU),
related information across fast changing networks. More- random access memory (RAM), Graphics Processing Unit
over, in case of constrained network, it become very difficult (GPU) etc.) [69].
to implement complex security mechanism. Open research
challenges are the design of schemes that would deal with G. MOBILITY
powerful and also constrained devices in order to ensure The architecture for ICN based Edge computing should sup-
privacy and integrity. port consumer mobility and producer mobility as well. Since
the consumer mobility is inherently supported in ICN, how-
E. NAMING ever, there is much work needed for the producer mobility.
Naming is also an open issue for ICN over Edge networks. Only limited number of works is available dealing with pro-
The use of ICN in Edge computing creates issues, since ducer mobility such as [114]–[116]. However, most of these
ICN access the data and services by name. There is a need works cover fixed networks. Producer mobility needs to be
to introduce the design of novel naming schemes that sup- addressed for Edge computing, since in ICN based Edge
port both Edge and ICN to handle mobility, security, and system, the mobile users themselves might be publishers to
scalability. To be more specific, content and context-based a local audience, supported by Edge nodes for storage and
Naming Schemes could be designed for ICN based Edge computing capacity. Therefore, those mobile devices may
systems [112]. Where context-based naming means the infor- behave like a publisher/producer and may move in the entire
mation on the context i-e location (where are the users) iden- network. Mechanisms are required to address the producer
tity (who are the users), neighbors (who are near them) and mobility of ICN based Edge system.
what are the content choices and so on. In ICN the location,
identity, neighbor etc) could be named and signed pieces of H. NETWORK HETEROGENEITY
content that will reflect that local context. The named content Future networks will support numerous wireless technologies
could be relevant object(s) that may be needed sooner or later. and types of devices. Therefore, a mobile user can move
In addition, from a consumer perspective, there are mul- across diverse networks. However, this mobility will result in
tiple options to request for data such as query for data handoff delay, packet loss, packet duplication, or packet re-
objects or chunks using their name or sequence number. ordering for the period of handoff. This handoff is an issue
When talking about services, querying for results (e.g. in such heterogeneous networks. Future networks such as
function results) becomes difficult. Customized informa- 5G and beyond will need a minimum handoff delay when-
tion or parameters need to be provided by a consumer, for ever mobile users move across different networks. Therefore,
example as part of the naming scheme such as the NFN ICN over Edge computing is an efficient approach to deal
approach. Research activities need to investigate the options with such issues. Now the question arises: How ICN over
to querying the network for computational expensive and Edge computing can be an efficient solution? To answer this,

73484 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

we explain features of each paradigm and how they will help big data management, mobility, naming, latency, and security.
to solve these issues. In Edge computing, data processing is In this article, we provided a comprehensive survey of ICN
pushed from the Cloud to the Edge of the network to achieve over Edge computing for future networks. The contributions
minimum latency. Thus, Edge computing offers less delay are multi-fold. First, the Edge computing concepts, drivers,
than the Cloud. 5G networks basic requirement is a minimum proposals, limitations and use cases has been provided. Sec-
delay, Therefore, this delay can be reduced more if we apply ond, ICN is highlighted for Edge computing which includes
ICN in Edge computing. How can ICN help? First, ICN is an overview of ICN and motivations to leverage the ICN for
receiver-driven in nature and supports consumer-mobility: Edge computing. In addition, various standardization efforts
unsatisfied requests will be re-issued whenever consumers and software and hardware vendor solutions are presented.
move. Second, host multi-homing is also supported by ICN. Finally, future research directions are provided for ICN over
Therefore, content requests and data delivery can use any Edge computing. We believe this survey will stimulate the
of the interface(s) available at the device [78]. Due to the research community and pave the way towards the empower-
disconnection of consumers and producers, self-consistent ment of future networks in order to achieve the fast response
content, any casting, and in-network-caching, ICN thus time.
proves to reconnect services and devices in heterogeneous
networks [21], [69]. In addition, function in ICN is named REFERENCES
that could be requested at the network layer. Explicitly named [1] A. U. R. Khan, M. Othman, S. A. Madani, and S. U. Khan, ‘‘A survey
of mobile cloud computing application models,’’ IEEE Commun. Surveys
functions can be resolved in network nodes, while network- Tuts., vol. 16, no. 1, pp. 393–413, 1st Quart., 2013.
layer requests (i.e., Interests) can carry input information for [2] M. Amadeo, C. Campolo, A. Molinaro, and G. Ruggeri, ‘‘IoT data pro-
Edge-executable functions [118]. We explained the features cessing at the edge with named data networking,’’ in Proc. Eur. Wireless
Conf., May 2018, pp. 1–6.
of both ICN and Edge computing above; therefore, joint [3] G. Li, J. Wu, J. Li, K. Wang, and T. Ye, ‘‘Service popularity-based smart
optimization will solve the issues of network heterogeneity, resources partitioning for fog computing-enabled industrial Internet of
mobility, and so on. Things,’’ IEEE Trans. Ind. Informat., vol. 14, no. 10, pp. 4702–4711,
Oct. 2018.
I. LOAD BALANCING STRATEGIES FOR REQUESTS ON [4] J. Wu, M. Dong, K. Ota, J. Li, and Z. Guan, ‘‘Big data analysis-based
secure cluster management for optimized control plane in software-
MULTI-INTERFACES defined networks,’’ IEEE Trans. Netw. Serv. Manage., vol. 15, no. 1,
For an efficient content retrieval in ICN based Edge comput- pp. 27–38, Mar. 2018.
[5] G. Li, J. Wu, J. Li, T. Ye, and R. Morello, ‘‘Battery status sensing
ing, the network stack must transparently support simulta- software-defined multicast for V2G regulation in smart grid,’’ IEEE
neous use of multiple interfaces. In TCP/IP this is done via Sensors J., vol. 17, no. 23, pp. 7838–7848, Dec. 2017.
multipath TCP [117], [118]. The problem in multipath TCP [6] A. C. Baktir, A. Ozgovde, and C. Ersoy, ‘‘How can edge computing
benefit from software-defined networking: A survey, use cases, and future
is that we should know in advance that on how many paths directions,’’ IEEE Commun. Surveys Tuts., vol. 19, no. 4, pp. 2359–2391,
we want to load balance our requests. However, ICN provides 4th Quart., 2017.
native support for multi-interface communication. Therefore, [7] M. Ahmadi, N. Khanezaei, S. Manavi, F. F. Moghaddam, and
T. Khodadadi, ‘‘A comparative study of time management and energy
load balancing strategies are required for ICN based Edge
consumption in mobile cloud computing,’’ in Proc. IEEE 5th Control Syst.
computing that would try to minimize the congestion over the Graduate Res. Colloq., Aug. 2014, pp. 199–203.
entire network. Moreover, how to tune the request on different [8] B. Zhou, A. V. Dastjerdi, R. N. Calheiros, S. N. Srirama, and R. Buyya,
interfaces without degrading the user experience is an issue ‘‘A context sensitive offloading scheme for mobile cloud computing
service,’’ in Proc. IEEE 8th Int. Conf. Cloud Comput., New York, NY,
in ICN based Edge computing. USA, Jun. 2015, pp. 869–876.
[9] Y. Li, L. Sun, and W. Wang, ‘‘Exploring device-to-device communication
J. INFRASTRUCTURE AND ARCHITECTURE for mobile cloud computing,’’ in Proc. IEEE Int. Conf. Commun. (ICC),
We believe that the combination of ICN and Edge comput- Sydney, NSW, Australia, Jun. 2014, pp. 2239–2244.
[10] K. Ha et al., ‘‘The impact of mobile multimedia applications on data
ing will speed up content retrieval. However, ICN in Edge center consolidation,’’ in Proc. IEEE Int. Conf. Cloud Eng. (IC2E),
networks poses many challenges to network’s infrastructure Redwood City, CA, USA, Mar. 2013, pp. 166–176.
and architecture. The issue arises because of two different [11] M. T. Beck et al., ‘‘ME-VoLTE: Network functions for energy-efficient
video transcoding at the mobile edge,’’ in Proc. 18th IEEE Int. Conf.
architectures. How will ICN and Edge computing co-exist Intell. Next Generat. Netw., Paris, France, Apr. 2015, pp. 38–44.
with each other? Inter-networking schemes with existing [12] T. J. Barnett, A. Sumits, S. Jain, and U. Andra, ‘‘Cisco visual networking
architecture of Edge computing are necessary to make both index: Global mobile data traffic forecast update, 2016–2021,’’ Cisco
Syst., Inc., San Jose, CA, USA, White Paper 1454457600, Mar. 2017.
paradigms interoperable. The interoperability of ICN with [Online]. Available: http://www.cisco.com/c/en/us/solutions/collateral/
Edge networks is an open challenge to researchers that should service-provider/visual-networking-index-vni/mobile-white-paper-c11-
be addressed. 520862.html
[13] M. T. Beck et al., ‘‘Mobile edge computing: A taxonomy,’’ in Proc. 6th
XII. CONCLUSION Int. Conf. Adv. Future Internet, Lisbon, Portugal, 2014, pp. 48–54.
[14] M. Satyanarayanan, ‘‘The emergence of edge computing,’’ Computer,
A migration of IP addresses and the use of naming mecha- vol. 50, no. 1, pp. 30–39, Jan. 2017.
nisms for content, context and services/functions shall con- [15] F. Bonomi, R. Milito, P. Natarajan, and J. Zhu, ‘‘Fog computing:
tribute to the optimal performance for future networks. ICN A platform for Internet of Things and analytics,’’ in Big Data and
Internet of Things: A Roadmap for Smart Environments. Springer,
over Edge computing is an efficient technique to ensure a Mar. 2014, pp. 169–186. [Online]. Available: https://link.springer.com/
shorter response time. It will help to solve the problems of chapter/10.1007/978-3-319-05029-4_7

VOLUME 6, 2018 73485

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

[16] D. Thomas, ‘‘Cloud computing—Benefits and challenges!’’ J. Object [40] C. Byers. OpenFog Consortium—Out of the Fog: Use Case
Technol., vol. 8, no. 3, pp. 37–41, Jun. 2009. Scenarios. Accessed: May 5, 2017. [Online]. Available:
[17] L. Gao, T. H. Luan, B. Liu, W. Zhou, and S. Yu, ‘‘Fog com- https://www.openfogconsortium.org/wp-content/uploads/OpenFog-
puting and its applications in 5G,’’ in 5G Mobile Communications. Transportation-Drone-Delivery-Use-Case.pdf
Cham, Switzerland: Springer, 2017, pp. 571–593. [Online]. Available: [41] M. Aazam and E.-N. Huh, ‘‘Fog computing and smart gateway based
https://doi.org/10.1007/978-3-319-34208-5_21 communication for cloud of things,’’ in Proc. IEEE Int. Conf. Future
[18] Y. Yu, ‘‘Mobile edge computing towards 5G: Vision, recent progress, Internet Things Cloud, Barcelona, Spain, Aug. 2014, pp. 464–470.
and open challenges,’’ China Commun., vol. 13, no. 2, pp. 89–99, [42] M. Firdhous, O. Ghazali, and S. Hassan, ‘‘Fog computing: Will it be
2016. the future of cloud computing?’’ in Proc. 3rd Int. Conf. Inform. Appl.,
[19] M. Paoloni, ‘‘Power at the edge: Processing and storage move from Kuala Terengganu, Malaysia, 2014, pp. 8–15.
Central core to the network edge,’’ Greater Seattle Area, Senza Fili [43] T. H. Luan, L. Gao, Z. Li, Y. Xiang, G. Wei, and L. Sun. (2015). ‘‘Fog
Consulting, Tech. Rep., Feb. 2017. [Online]. Available: http://www. computing: Focusing on mobile users at the edge.’’ [Online]. Available:
senzafiliconsulting.com https://arxiv.org/abs/1502.01815
[20] (Jun. 2017). Ericsson Mobility Report, SE-164 80 Stockholm, Sweden. [44] P. Hu, H. Ning, T. Qiu, Y. Zhang, and X. Luo, ‘‘Fog computing based face
[Online]. Available: https://www.ericsson.com/en/mobility-report identification and resolution scheme in Internet of Things,’’ IEEE Trans.
[21] K. Gai, M. Qiu, H. Zhao, L. Tao, and Z. Zong, ‘‘Dynamic energy-aware Ind. Informat., vol. 13, no. 4, pp. 1910–1920, Aug. 2017.
cloudlet-based mobile cloud computing model for green computing,’’ [45] E. Cau et al., ‘‘Efficient exploitation of mobile edge computing for
J. Netw. Comput. Appl., vol. 59, pp. 46–54, Jan. 2016. virtualized 5G in EPC architectures,’’ in Proc. 4th IEEE Int. Conf.
[22] S. Marston, Z. Li, S. Bandyopadhyay, J. Zhang, and A. Ghalsasi, ‘‘Cloud Mobile Cloud Comput., Services, Eng. (Mobile Cloud), Oxford, U.K.,
computing—The business perspective,’’ Decision Support Syst., vol. 51, Mar./Apr. 2016, pp. 100–109.
no. 1, pp. 176–189, 2011. [46] M. Agiwal, A. Roy, and N. Saxena, ‘‘Next generation 5G wireless net-
[23] M. Rahman, J. Gao, and W.-T. Tsai, ‘‘Energy saving in mobile cloud works: A comprehensive survey,’’ IEEE Commun. Surveys Tut., vol. 18,
computing,’’ in Proc. IEEE Int. Conf. Cloud Eng. (IC2E), San Francisco, no. 3, pp. 1617–1655, 3rd Quart., 2016.
CA, USA, Mar. 2013, pp. 285–291. [47] S. Nunna et al., ‘‘Enabling real-time context-aware collaboration through
[24] Z. Sanaei, S. Abolfazli, A. Gani, and R. Buyya, ‘‘Heterogeneity 5G and mobile edge computing,’’ in Proc. 12th Int. Conf. Inf. Technol.-
in mobile cloud computing: Taxonomy and open challenges,’’ New Generat., Las Vegas, NV, USA, Apr. 2015, pp. 601–605.
IEEE Commun. Surveys Tuts., vol. 16, no. 1, pp. 369–392,
[48] D. Satria, D. Park, and M. Jo, ‘‘Recovery for overloaded mobile edge
Feb. 2014.
computing,’’ Future Generat. Comput. Syst., vol. 70, pp. 138–147,
[25] M. Satyanarayanan et al., ‘‘Edge analytics in the Internet of Things,’’ May 2017.
IEEE Pervasive Comput., vol. 14, no. 2, pp. 24–31, Feb. 2015.
[49] European Telecommunications Standards Institute Industry Specifica-
[26] R. Roman, J. Lopez, and M. Mambo, ‘‘Mobile edge computing, Fog et al.: tions Group, Mobile-Edge Computing—Service Scenarios. [Online].
A survey and analysis of security threats and challenges,’’ Future Generat. Available: http://www.etsi.org/deliver/etsi_gs/MEC-IEG/001_099/004/
Comput. Syst., vol. 78, pp. 680–698, Nov. 2016. [Online]. Available: 01.01.01_60/gs_MEC-IEG004v010101p.pdf
https://doi.org/10.1016/j.future.2016.11.009
[50] M. Satyanarayanan et al., ‘‘An open ecosystem for mobile-cloud conver-
[27] M. Satyanarayanan, Z. Chen, K. Ha, W. Hu, W. Richter, and
gence,’’ IEEE Commun. Mag., vol. 53, no. 3, pp. 63–70, Mar. 2015.
P. Pillai, ‘‘Cloudlets: At the leading edge of mobile-cloud convergence,’’
in Proc. 6th Int. Conf. Mobile Comput., Appl. Services, Austin, TX, USA, [51] A. Ahmed and E. Ahmed, ‘‘A survey on mobile edge computing,’’ in
Nov. 2014, pp. 1–9. Proc. 10th IEEE Int. Conf. Intell. Syst. Control ISCO, Coimbatore, India,
Jan. 2016, pp. 1–8.
[28] U. Shaukat, E. Ahmed, Z. Anwar, and F. Xia, ‘‘Cloudlet deployment in
local wireless networks: Motivation, architectures, applications, and open [52] M. Quwaider and Y. Jararweh, ‘‘Cloudlet-based efficient data collection
challenges,’’ J. Netw. Comput. Appl., vol. 62, pp. 18–40, Feb. 2016. in wireless body area networks,’’ Simul. Model. Pract. Theory, vol. 50,
[29] B. Li, Y. Pei, H. Wu, and B. Shen, ‘‘Heuristics to allocate high- pp. 57–71, Jan. 2015.
performance cloudlets for computation offloading in mobile ad hoc [53] S. Yi, Z. Hao, Z. Qin, and Q. Li, ‘‘Fog computing: Platform and appli-
clouds,’’ J. Supercomput., vol. 71, no. 8, pp. 3009–3036, 2015. cations,’’ in Proc. 3rd IEEE Workshop Hot Topics Web Syst. Technol.
[30] X. Guo, L. Liu, Z. Chang, and T. Ristaniemi, ‘‘Data offloading and task (HotWeb), Washington, DC, USA, Nov. 2015, pp. 73–78.
allocation for cloudlet-assisted ad hoc mobile clouds,’’ Wireless Netw., [54] P. Du and A. Nakao, ‘‘Application specific mobile edge computing
vol. 24, no. 1, pp. 79–88, Jan. 2018, doi: 10.1007/s11276-016-1322-z. through network softwarization,’’ in Proc. 5th IEEE Int. Conf. Cloud
[31] Open Edge Computing. Accessed: May 11, 2017. [Online]. Available: Netw. (Cloudnet), Pisa, Italy, Oct. 2016, pp. 130–135.
http://openedgecomputing.org/index.html [55] European Telecommunications Standards Institute Industry Specifica-
[32] A. Ceselli, M. Premoli, and S. Secci, ‘‘Cloudlet network design opti- tions Group, MEC Proofs of Concept. Accessed: May 15, 2017. [Online].
mization,’’ in Proc. IFIP Netw. Conf. (IFIP Netw.), Toulouse, France, Available: http://www.etsi.org/technologies-clusters/technologies/
May 2015, pp. 1–9. mobile-edge-computing/mec-poc
[33] European Telecommunications Standards Institute. Accessed: [56] European Telecommunications Standards Institute, PoC Edge Video
Sep. 10, 2018. [Online]. Available: http://www.etsi.org/ Orchestration and Video Clip Replay. Accessed: May 15, 2017. [Online].
[34] M. R. Rahimi, J. Ren, C. H. Liu, A. V. Vasilakos, and Available: http://mecwiki.etsi.org/index.php?title=PoC_2_Edge_Video_
N. Venkatasubramanian, ‘‘Mobile cloud computing: A survey, state Orchestration_ and _Video_Clip_Replay_via_MEC
of art and future directions,’’ Mobile Netw. Appl., vol. 19, no. 2, [57] K. Ha, Z. Chen, W. Hu, W. Richter, P. Pillai, and M. Satyanarayanan,
pp. 133–143, 2014. ‘‘Towards wearable cognitive assistance,’’ in Proc. 12th Annu. Int. Conf.
[35] W. Shi, J. Cao, Q. Zhang, Y. Li, and L. Xu, ‘‘Edge computing: Vision Mobile Syst., Appl., Services, New York, NY, USA, 2014, pp. 68–81.
and challenges,’’ IEEE Internet Things J., vol. 3, no. 5, pp. 637–646, [58] O. Arias, J. Wurm, K. Hoang, and Y. Jin, ‘‘Privacy and security in Internet
Oct. 2016. of Things and wearable devices,’’ IEEE Trans. Multi-Scale Comput. Syst.,
[36] B. Ahlgren, C. Dannewitz, C. Imbrenda, D. Kutscher, and B. Ohlman, vol. 1, no. 2, pp. 99–109, Apr./Jun. 2015.
‘‘A survey of information-centric networking,’’ IEEE Commun. Mag., [59] M. Quwaider and Y. Jararweh, ‘‘Cloudlet-based for big data collection
vol. 50, no. 7, pp. 26–36, Jul. 2012. in body area networks,’’ in Proc. 8th Int. Conf. Internet Technol. Secured
[37] F. Bonomi, R. Milito, J. Zhu, and S. Addepalli, ‘‘Fog computing and its Trans., London, U.K., Dec. 2013, pp. 137–141.
role in the Internet of Things,’’ in Proc. 1st Ed. MCC Workshop Mobile [60] Y. Shi, G. Ding, H. Wang, H. E. Roman, and S. Lu, ‘‘The fog comput-
Cloud Comput., Helsinki, Finland, 2012, pp. 13–16. ing service for healthcare,’’ in Proc. 2nd Int. Symp. Future Inf. Com-
[38] I. Stojmenovic and S. Wen, ‘‘The Fog computing paradigm: Scenarios and mun. Technol. Ubiquitous HealthCare (Ubi-HealthTech), Beijing, China,
security issues,’’ in Proc. Federated Conf. Comput. Sci. Inf. Syst., Warsaw, May 2015, pp. 1–5.
Poland, Sep. 2014, pp. 1–8. [61] Q. Yaseen, F. AlBalas, Y. Jararweh, and M. Al-Ayyoub, ‘‘A fog comput-
[39] OpenFog Consortium, Fremont, CA, USA. An Open, Interoperable Fog ing based system for selective forwarding detection in mobile wireless
Computing Architecture. Accessed: May 5, 2017. [Online]. Available: sensor networks,’’ in Proc. IEEE 1st Int. Workshops Found. Appl. Self∗
https://www.openfogconsortium.org/about-us/ Syst. (FAS∗ W), Augsburg, Germany, Sep. 2016, pp. 256–262.

73486 VOLUME 6, 2018

R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

[62] M. Satyanarayanan, G. Lewis, E. Morris, S. Simanta, J. Boleng, and [86] R. El Hattachi and J. Erfanian, ‘‘A Deliverable by NGMN (Next Genera-
K. Ha, ‘‘The role of cloudlets in hostile environments,’’ IEEE Pervas. tion Mobile Networks),’’ NGMN 5G, Frankfurt, Germany, White Paper,
Comput., vol. 12, no. 4, pp. 40–49, Oct. 2013. Feb. 2015, pp. 1–125.
[63] V. S. Achanta, N. T. Sureshbabu, V. Thomas, M. L. Sahitya, and [87] E. Bastug, M. Bennis, and M. Debbah, ‘‘Living on the edge: The role
S. Rao, ‘‘Cloudlet-based multi-lingual dictionaries,’’ in Proc. 3rd Int. of proactive caching in 5G wireless networks,’’ IEEE Commun. Mag.,
Conf. Services Emerg. Markets, Mysore, India, Dec. 2012, pp. 30–36. vol. 52, no. 8, pp. 82–89, Aug. 2014.
[64] R. Ullah, Y. Faheem, and B. S. Kim, ‘‘Energy and congestion-aware [88] M. Simsek, A. Aijaz, M. Dohler, J. Sachs, and G. Fettweis, ‘‘5G-enabled
routing metric for smart grid ami networks in smart city,’’ IEEE Access, tactile Internet,’’ IEEE J. Sel. Areas Commun., vol. 34, no. 3, pp. 460–473,
vol. 5, pp. 13799–13810, 2017, doi: 10.1109/ACCESS.2017.2728623. Mar. 2016.
[65] I. Stojmenovic, ‘‘Fog computing: A cloud to the ground support for smart [89] O. Galinina, A. Pyattaev, S. Andreev, M. Dohler, and Y. Koucheryavy,
things and machine-to-machine networks,’’ in Proc. Austral. Telecom- ‘‘5G multi-RAT LTE-WiFi ultra-dense small cells: Performance dynam-
mun. Netw. Appl. Conf. (ATNAC), Southbank, VIC, Australia, Nov. 2014, ics, architecture, and trends,’’ IEEE J. Sel. Areas Commun., vol. 33, no. 6,
pp. 117–122. pp. 1224–1240, Jun. 2015.
[66] Y. Jararweh, A. Doulat, O. AlQudah, E. Ahmed, M. Al-Ayyoub, and [90] M. Amadeo et al., ‘‘Information-centric networking for the Internet
E. Benkhelifa, ‘‘The future of mobile cloud computing: Integrating of Things: Challenges and opportunities,’’ IEEE Netw., vol. 30, no. 2,
cloudlets and mobile edge computing,’’ in Proc. 23rd Int. Conf. Telecom- pp. 92–100, Mar./Apr. 2016.
mun. (ICT), Thessaloniki, Greece, May 2016, pp. 1–5. [91] D. Trossen, M. J. Reed, M. Georgiades, N. Fotiou, and
[67] M. Aazam and E.-N. Huh, ‘‘Fog computing: The cloud-IoT/IoE mid- G. Xylomenos. (2015). ‘‘IP over ICN–The better IP? An unusual take on
dleware paradigm,’’ IEEE Potentials, vol. 35, no. 3, pp. 40–44, information-centric networking.’’ [Online]. Available: https://arxiv.org/
May/Jun. 2016. abs/1507.04221
[68] Y. Liu, X. Mao, Y. He, K. Liu, W. Gong, and J. Wang, ‘‘CitySee: Not [92] M. Amadeo et al., ‘‘Information-centric networking for the Internet
only a wireless sensor network,’’ IEEE Netw., vol. 27, no. 5, pp. 42–47, of Things: Challenges and opportunities,’’ IEEE Netw., vol. 30, no. 2,
Oct. 2013. pp. 92–100, Mar./Apr. 2016.
[69] L. Soh, J. Burke, and L. Zhang, ‘‘Supporting augmented reality: Look- [93] W. Shi, J. Cao, Q. Zhang, Y. Li, and L. Xu, ‘‘Edge computing: Vision
ing beyond performance,’’ in Proc. Morning Workshop Virtual Reality and challenges,’’ IEEE Internet Things J., vol. 3, no. 5, pp. 637–646,
Augmented Reality Netw. (VR/AR Netw.), New York, NY, USA, 2018, Oct. 2016.
pp. 7–12, doi: 10.1145/3229625.3229627. [94] X. Wang, M. Chen, T. Taleb, A. Ksentini, and V. C. M. Leung,
[70] J. Grubert, T. Langlotz, S. Zollmann, and H. Regenbrecht, ‘‘Towards ‘‘Cache in the air: Exploiting content caching and delivery techniques
pervasive augmented reality: Context-awareness in augmented reality,’’ for 5G systems,’’ IEEE Commun. Mag., vol. 52, no. 2, pp. 131–139,
IEEE Trans. Vis. Comput. Graphics, vol. 23, no. 6, pp. 1706–1724, Feb. 2014.
Jun. 2017. [95] G. A. Lewis, S. Echevería, S. Simanta, B. Bradshaw, and
[71] G. Karagiannis et al., ‘‘Vehicular networking: A survey and tutorial on J. Root, ‘‘Cloudlet-based cyber-foraging for mobile systems in resource-
requirements, architectures, challenges, standards and solutions,’’ IEEE constrained edge environments,’’ in Proc. 36th Int. Conf. Softw. Eng.,
Commun. Surveys Tuts., vol. 13, no. 4, pp. 584–616, Nov. 2011. Hyderabad, India, May/Jun. 2014, pp. 412–415.
[72] M. Amadeo, C. Campolo, and A. Molinaro, ‘‘Information-centric net- [96] Extreme outdoor Server. ADLINK Technologies.
working for connected vehicles: A survey and future perspectives,’’ IEEE Accessed: Jul. 15, 2018. [Online]. Available: http://www.adlinktech.com/
Commun. Mag., vol. 54, no. 2, pp. 98–104, 4th Quart., 2016. PD/web/PD_detail.php?cKind=&pid=1573&utm_source
[73] M. Kaur, J. Martin, and H. Hu, ‘‘Comprehensive view of security prac- [97] M. Paolini and S. Fili, ‘‘How to control latency and benefit
tices in vehicular networks,’’ in Proc. Int. Conf. Connected Vehicles from it: An expanded role for latency,’’ Greater Seattle Area,
Expo (ICCVE), Seattle, WA, USA, Sep. 2016, pp. 19–26. Senza Fili Consulting, White Paper, 2018, [Online]. Available:
[74] A. Osseiran et al., ‘‘Scenarios for 5G mobile and wireless communica- http://www.senzafiliconsulting.com
tions: The vision of the METIS project,’’ IEEE Commun. Mag., vol. 52, [98] D. Grewe, M. Wagner, M. Arumaithurai, I. Psaras, and D. Kutscher,
no. 5, pp. 26–35, May 2014. ‘‘Information-centric mobile edge computing for connected vehicle envi-
[75] M. F. Bari, S. R. Chowdhury, R. Ahmed, R. Boutaba, and B. Mathieu, ronments: Challenges and research directions,’’ in Proc. Workshop Mobile
‘‘A survey of naming and routing in information-centric networks,’’ IEEE Edge Commun. (MECOMM), New York, NY, USA, 2017, pp. 7–12, doi:
Commun. Mag., vol. 50, no. 12, pp. 44–53, Dec. 2012. 10.1145/3098208.3098210.
[76] G. Xylomenos et al., ‘‘A survey of information-centric networking [99] N. Fotiou, V. A. Siris, G. Xylomenos, G. C. Polyzos, K. V. Katsaros, and
research,’’ IEEE Commun. Surveys Tuts., vol. 16, no. 2, pp. 1024–1049, G. Petropoulos, ‘‘Edge-ICN and its application to the Internet of Things,’’
2nd Quart., 2014. in Proc. IFIP Netw. Conf. (IFIP Netw.) Workshops, Stockholm, Sweden,
[77] C. Fang, F. R. Yu, T. Huang, J. Liu, and Y. Liu, ‘‘A survey of green Jun. 2017, pp. 1–6, doi: 10.23919/IFIPNetworking.2017.8264880.
information-centric networking: Research issues and challenges,’’ IEEE [100] M. Sifalakis, B. Kohler, C. Scherb, and C. Tschudin, ‘‘An information
Commun. Surveys Tuts., vol. 17, no. 3, pp. 1455–1472, 3rd Quart., 2015. centric network for computing the distribution of computations,’’ in Proc.
[78] M. Amadeo, C. Campolo, and A. Molinaro, ‘‘Information-centric net- 1st ACM Conf. Inf.-Centric Netw. (ACM-ICN), New York, NY, USA,
working for connected vehicles: A survey and future perspectives,’’ IEEE 2014, pp. 137–146, doi: 10.1145/2660129.2660150.
Commun. Mag., vol. 54, no. 2, pp. 98–104, Feb. 2016. [101] Docker Inc. (2017). The Docker Project Page. [Online]. Available:
[79] M. Satyanarayanan, P. Bahl, R. Caceres, and N. Davies, ‘‘The case for https://www.docker.com/
VM-based cloudlets in mobile computing,’’ IEEE Pervasive Comput., [102] Amazon Web Services, Inc. (2017). Amazon Lambda Project Page—
vol. 8, no. 4, pp. 14–23, Oct./Dec. 2009. Run Code, Not Servers—Serverless Computing. [Online]. Available:
[80] Cloudlet-based Edge Computing. Accessed: May 17, 2017. [Online]. https://aws.amazon.com/lambda/
Available: http://elijah.cs.cmu.edu/ [103] A. Madhavapeddy, ‘‘Unikernels: Library operating systems for the
[81] 3rd Generation Partnership Project (3GPP). [Online]. Available: cloud,’’ SIGARCH Comput. Archit. News, vol. 41, no. 1, pp. 461–472,
http://www.3gpp.org/ 2013.
[82] Information-Centric Networking Research Group ICNRG. [Online]. [104] W. Shang et al., ‘‘Named data networking of things (invited paper),’’ in
Available: https://irtf.org/icnrg Proc. IEEE 1st Int. Conf. Internet-Things Design Implement. (IoTDI),
[83] SG13: Future Networks, With focus on IMT-2020, Cloud Computing Berlin, Germany, Apr. 2016, pp. 117–128, doi: 10.1109/IoTDI.2015.44.
and Trusted Network Infrastructures. [Online]. Available: http://www. [105] M. Mukherjee, D. Wang, and L. Shu, ‘‘Survey of fog computing: Funda-
itu.int/en/ITU-T/studygroups/2017-2020/13/Pages/default.asp mental, network applications, and research challenges,’’ IEEE Commun.
[84] Alliance for Telecommunications Industry Solutions. Surveys Tuts, vol. 20, no. 3, pp. 1826–1857, 3rd Quart., 2018.
Accessed: May 15, 2017. [Online]. Available: https://www.atis.org/ [106] ADLINK Technology. Accessed: Jul. 15, 2018. [Online]. Available:
[85] M. Recchione, ‘‘Understanding information centric networking and https://www.adlinktech.com
mobile edge computing,’’ 5G Americas, Vicki Livingston, White [107] Advantech. [Online]. Available: www.advantech.com/nc
Paper 3330, Dec. 2016. [Online]. Available: http://www.5gamericas.org/ [108] Artesyn. [Online]. Available: https://www.artesyn.com/
files/3414/8173/2353/Understanding_Information_Centric_Networking_ [109] Interdigital. [Online]. Available: http://www.interdigital.com/
and_Mobile_Edge_Computing.pdf [110] Qwilt. [Online]. Available: https://www.qwilt.com/

VOLUME 6, 2018 73487

 R. Ullah et al.: ICN With Edge Computing for IoT: Research Challenges and Future Directions

[111] Vasona Networks. Accessed: Jul. 15, 2018. [Online]. Available: SYED HASSAN AHMED (S’13–M’17–SM’18)
https://www.vasonanetworks.com/ received the B.S. degree in computer science from
[112] K. Zhang, Y. Mao, S. Leng, Y. He, and Y. Zhang, ‘‘Mobile-edge com- the Kohat University of Science and Technology,
puting for vehicular networks: A promising network paradigm with pre- Pakistan, and the master’s combined Ph.D. degree
dictive off-loading,’’ IEEE Veh. Technol. Mag., vol. 12, no. 2, pp. 36–44, from the School of Computer Science and Engi-
Jun. 2017. neering (SCSE), Kyungpook National University
[113] S. Arshad, B. Shahzaad, M. A. Azam, J. Loo, S. H. Ahmed, and (KNU), South Korea. In 2015, he was a Visiting
S. Aslam, ‘‘Hierarchical and flat-based hybrid naming scheme in content-
Ph.D. Student at Georgia Tech, Atlanta, GA, USA.
centric networks of things,’’ IEEE Internet Things J., vol. 5, no. 2,
He is currently an Assistant Professor with the
pp. 1070–1080, Apr. 2018.
[114] T. Refaei, J. Ma, S. Ha, and S. Liu, ‘‘Integrating IP and NDN through Computer Science Department, Georgia Southern
an extensible IP-NDN gateway,’’ in Proc. ACM ICN, vol. 3, 2017, University (GSU), Statesboro, GA, USA. Before starting at GSU, he was
pp. 224–225, a Post-Doctoral Fellow with the Department of Electrical and Computer
[115] J. Augé, G. Carofiglio, G. Grassi, L. Muscariello, G. Pau, and X. Zeng, Engineering, University of Central Florida, Orlando, FL, USA. He has
‘‘MAP-Me: Managing anchor-less producer mobility in content-centric authored/co-authored over 100 international publications, including jour-
networks,’’ IEEE Trans. Netw. Serv. Manage., vol. 15, no. 2, pp. 596–610, nal articles, conference proceedings, book chapters, and three books. His
Jun. 2018, doi: 10.1109/TNSM.2018.2796720. research interests include sensor and ad hoc networks, cyber-physical sys-
[116] Y. Zhang, A. Afanasyev, J. Burke, and L. Zhang, ‘‘A survey of mobil- tems, vehicular communications, and future Internet. He is an ACM Profes-
ity support in named data networking,’’ in Proc. 3rd Workshop Name- sional Member. From 2014 to 2016, he consequently received the Research
Oriented Mobility, Architect., Algorithms Appl. (NOM), Apr. 2016, Contribution Awards by SCSE and KNU. In 2016, his work on robust content
pp. 83–88. retrieval in future vehicular networks leads him to receive the Qualcomm
[117] Z. Zhu, R. Wakikawa, and L. Zhang, A Survey of Mobility Support in the Innovation Award at KNU.
Internet. document RFC 6301.IRTF, 2011.
[118] C.-A. Sarros et al., ‘‘Connecting the edges: A universal, mobile-centric,
and opportunistic communications architecture,’’ IEEE Commun. Mag.,
vol. 56, no. 2, pp. 136–143, Feb. 2018.
[119] X. Chen, L. Jiao, W. Li, and X. Fu, ‘‘Efficient multi-user computation
offloading for mobile-edge cloud computing,’’ IEEE/ACM Trans. Netw.,
vol. 24, no. 5, pp. 2795–2808, Oct. 2016.
[120] Z. Yantao and M. Jianfeng, ‘‘A highly secure identity-based authenticated
key-exchange protocol for satellite communication,’’ J. Commun. Netw.,
vol. 12, no. 6, pp. 592–599, Dec. 2011.
[121] S. S. Al-Riyami and K. G. Paterson, ‘‘Certificateless public key cryp-
tography,’’ in Advances in Cryptology—ASIACRYPT, vol. 2894. Berlin,
Germany: Springer-Verlag, 2003, pp. 452–473.
[122] J. Wu, M. Dong, K. Ota, J. Li, and Z. Guan, ‘‘FCSS: Fog computing based
content-aware filtering for security services in information centric social
networks,’’ IEEE Trans. Emerg. Topics Comput., to be published, doi:
10.1109/TETC.2017.2747158. BYUNG-SEO KIM (M’02–SM’17) received the
[123] Y. Liu, A. Zhang, J. Li, and J. Wu, ‘‘An anonymous distributed key B.S. degree in electrical engineering from Inha
management system based on CL-PKC for space information network,’’ University, Incheon, South Korea, in 1998, and
in Proc. IEEE Int. Conf. Commun. (ICC), Kuala Lumpur, Malaysia, the M.S. and Ph.D. degrees in electrical and com-
May 2016, pp. 1–7, doi: 10.1109/ICC.2016.7510841. puter engineering from the University of Florida
[124] M. Król and I. Psaras, ‘‘NFaaS: Named function as a service,’’ in Proc. in 2001 and 2004, respectively. His Ph.D. study
ACM Conf. Inf.-Centric Netw., Berlin, Germany, 2017, pp. 134–144.
was supervised by Dr. Y. Fang. From 1997 to
[125] M. Amadeo, C. Campolo, and A. Molinaro, ‘‘NDNe: Enhancing named
1999, he was at Motorola Korea Ltd., Paju, South
data networking to support cloudification at the edge,’’ IEEE Commun.
Lett., vol. 20, no. 11, pp. 2264–2267, Nov. 2016. Korea, as a Computer Integrated Manufacturing
Engineer in Advanced Technology Research and
Development. From 2005 to 2007, he was at Motorola Inc., Schaumburg
Illinois, as a Senior Software Engineer in Networks and Enterprises. His
REHMAT ULLAH received the B.S and M.S research focuses in Motorola Inc. were designing protocol and network
degrees (major in wireless communications and architecture of wireless broadband mission critical communications. From
networks) in computer science from COM- 2012 to 2014, he was the Chairman with the Department of Software and
SATS University Islamabad, Islamabad, Pakistan, Communications Engineering, Hongik University, South Korea, where he
in 2013 and 2016, respectively. He is currently is currently a Professor. His work has appeared in around 167 publications
pursuing the Ph.D. degree in computer engineering and 22 patents. His research interests include the design and development
with the Broadband Convergence Networks Lab- of efficient wireless/wired networks, including link-adaptable/cross-layer-
oratory, Department of Electronics and Computer based protocols, multi-protocol structures, wireless CCNs/NDNs, mobile
Engineering, Hongik University, South Korea. His edge computing, physical layer design for broadband PLC, and resource
major interests are in the field of information- allocation algorithms for wireless networks. He was also served as the
centric networking (specifically wireless content-centric/named data net- member of Sejong-city Construction Review Committee and the Ansan-city
working), edge computing, low power and lossy networks (LLN), IoT, 5th Design Advisory Board. He served as the General Chair for 3rd IWWCN
Generation (5G), and future internet architectures. Moreover, he is an ACM 2017, and a TPC member for the IEEE VTC 2014-Spring and the EAI
member and serves as a Reviewer for the IEEE Wireless Communications FUTURE2016, and ICGHIC in 2016 and 2019 conferences. He served as
Magazine, the IEEE Communications Magazine, Transactions on Emerging the Guest Editor of special issues for the International Journal of Distributed
Telecommunications Technologies, Future Generation Computer Systems Sensor Networks (SAGE), the IEEE ACCESS, and the Journal of the Institute of
(Elsevier), the IEEE ACCESS Journal, and various International Conferences Electrics and Information Engineers. He is an Associate Editor of the IEEE
and Workshops, including ACM MobiHoc (Los Angeles, CA, USA), IEEE ACCESS.
CCNC (Las Vegas, NV, USA), and VTC2018-Spring (Porto, Portugal).

73488 VOLUME 6, 2018

View publication stats