From the Sky to the Ground: Comparing Fog Computing with Related

Distributed Paradigms

Joao Bachiega Jr.

, Breno Costa

, Leonardo R. Carvalho

, Victor H. C. Oliveira

, William X. Santos

Maria Clicia S. de Castro

and Aleteia Araujo

Department of Computer Science, University of Bras

ılia (UnB), Brazil

Department of Infomatics and Computer Science, State University of Rio de Janeiro (UERJ), Rio de Janeiro, Brazil

wilxavier@me.com, clicia@ime.uerj.br, aleteia@unb.br

Keywords:

Edge Computing, Fog Computing, Cloud Computing, Sky Computing.

Abstract:

Fog computing is a paradigm that enables provisioning resources and services at the network edge, closer to

end devices, complementing cloud computing and recently it’s being embraced by sky computing. Beyond

this computational paradigm, there are many other related technologies which also make use of distributed

computing to improve the quality of service delivered to the end-users. The main contribution of this paper

is to present a comparison between fog computing and nine other relevant related paradigms, namely Sky

Computing, Cloud Computing, Edge Computing, Mobile Edge Computing, Mobile Cloud Computing, Mobile

Ad hoc Cloud Computing, Mist Computing, Cloudlet Computing, and Dew Computing, highlighting the

similarities and differences between them based on the main fog characteristics. A graphical characterization

for each paradigm, highlighting its computational power, communication type and position in a three-layers

architecture (Cloud-Fog-IoT) and some relevant challenges in this area are also presented.

1 INTRODUCTION

Over the years, computing paradigms have been

evolving. Since the beginning of the modern com-

puter era until about 1985, computers were large

and expensive. Moreover, there was no connectiv-

ity between them and they operated independently of

each other. With the development of powerful mi-

croprocessors and high-speed computer networks, it

was possible to change this scenario (Tanenbaum and

Van Steen, 2007).

After that, computing technologies evolved from

distributed, parallel, clustering, peer-to-peer (P2P),

and grid until the cloud computing. There is no

doubt that cloud computing is one of the most im-

portant computational paradigms, playing an essen-

tial role in almost every aspect of our everyday life. It

brought the impression of inﬁnite computational re-

sources (e.g. storage, processing) provided in a pay-

per-use basis, allowing fast scalability at a reasonable

cost. Nevertheless, as a centralized computing model,

computation happens in the cloud data centers, lo-

cated in speciﬁc places around the world, and far from

the end-users (Mukherjee et al., 2018).

As in a cycle, nowadays is the time when dis-

tributed computing paradigms are back in the spot-

light, aiming to overcome the limitations of central-

ized cloud computing processes. Internet of Things

(IoT) development demands this (Yousefpour et al.,

2019). Thus, new concepts were born ranging from

Sky Computing (Keahey et al., 2009), which is a

level above cloud computing and uses resources and

services from several providers simultaneously, un-

til Fog Computing (Bonomi et al., 2012), which is a

cloud close to the ground, addressing several issues of

connected devices, like high-bandwidth, geographical

dispersion, and low latency needs. Fog is both com-

plementary to, and an extension of, traditional cloud-

based models.

Although fog computing is a recent paradigm,

in the last 3 years it was the subject of approxi-

mately 20% of all publications related to computa-

tional paradigms, including Sky Computing, Edge

Computing, Mobile Edge Computing, Mobile Cloud

Computing, Mobile Ad hoc Cloud Computing, Mist

Computing, Cloudlet Computing and Dew Comput-

ing.

The main contribution of this paper is to present a

comparison between fog computing and other related

paradigms, highlighting similarities and differences

158

Bachiega Jr., J., Costa, B., Carvalho, L., Oliveira, V., Santos, W., S. de Castro, M. and Araujo, A.

From the Sky to the Ground: Comparing Fog Computing with Related Distributed Paradigms.

DOI: 10.5220/0011033300003200

In Proceedings of the 12th International Conference on Cloud Computing and Services Science (CLOSER 2022), pages 158-169

ISBN: 978-989-758-570-8; ISSN: 2184-5042

among them, based on the main fog characteristics.

However, as some deﬁnitions are still imprecise, even

nebulous, in the literature, we based our research on a

qualitative bibliographic approach. To present a more

accurate view of the whole computational spectrum,

from the Sky to the ground, we also provide a graph-

ical characterization for each paradigm, highlight-

ing its computational power, communication type and

position in the three-layers architecture (Cloud-Fog-

IoT). Besides this, important challenges will be pre-

sented and discussed, giving to the researchers a com-

prehensive view of the subject.

For that, this paper is organized as follows. Sec-

tions 2 and 3 presents the sky and cloud computing

concepts, respectively. In Section 4 the fog computing

characteristics are presented. The related paradigms

and technologies are described in Section 5, and a

comparison is presented and analyzed in Section 6.

The challenges in this area are presented in Section

7. Section 8 brings other papers related to this work.

Finally, Section 9 concludes this paper.

2 SKY COMPUTING

The market for cloud providers has become more

competitive over time. Currently, there is a profu-

sion of services offered by different providers. Many

of them offer similar products but using different ap-

proaches. In this scenario, a multicloud approach has

grown within organizations, in which resources of-

fered by different providers make up the technolog-

ical support framework of these organizations.

This approach has been called Sky computing

(Keahey et al., 2009) and can be deﬁned as a level

above cloud computing, since its resources are dy-

namically provisioned by different providers. It con-

sists of a layer of cloud environments’ management,

offering variable storage capacity with dynamic sup-

port for real-time demands.

Sky computing can also be found under the de-

nominations of multicloud (Kritikos and Plexousakis,

2015), cross-cloud (Elkhatib, 2016), federated clouds

(Paraiso et al., 2012), or inter-clouds (Grozev and

Buyya, 2014).This difference in nomenclature may be

related to the way the architectures deal with the re-

source scheduler, since this component can be intrin-

sic to the provider, or an external service provided by

a middleware, for example. There are also differences

regarding the form of integration between the clouds

participating in the sky computing layer, as well as re-

garding the knowledge of the existence of such layer

by the resources in their respective providers.

An important characteristic of a sky computing

environment is the consolidation of resources, whose

purpose is to offer a single view of all elements of

each cloud provider in the pool. To act in this con-

tinuous process of integration, discovery and consoli-

dation of resources, it is possible to use tools called

Cloud Orchestrators (de Carvalho and de Araujo,

2020) such as: Terraform (Shirinkin, 2017), Cloudify

(Cloudify, 2021), and Heat (Michelino et al., 2013),

for example. The sky computing characterization is

presented in Figure 1.

Figure 1: Sky Computing Characterization.

3 CLOUD COMPUTING

The underlying concept of cloud computing was in-

troduced in 1961 by John McCarthy when he said

that, in the future, computing could be organized as

a public service, as it was the telephone system in

those days (Foster et al., 2008). Computing and com-

munication evolution allowed the cloud computing

paradigm to expand and, nowadays, this platform is

largely used by academia and industries. Cloud com-

puting plays an important role as a computation in-

frastructure and impacts all other technologies and

paradigms presented in this paper.

We have observed, over the last few decades,

that computational resources, previously expensive

and scarce, have now become cheap and abundant

(Kushida et al., 2015). Cloud computing has emerged

in this transition context, enabling the computing de-

mocratization. It has encompassed two important

points: the illusion of inﬁnite computing resources,

and to allow the radical computation acceleration.

Cloud computing makes the concept of utility a re-

ality.

In the literature, we can ﬁnd in (Mell et al., 2011)

that the ﬁve essential features for cloud computing

environments are deﬁned as on-demand self-service,

broad network access, resource pooling, fast elastic-

ity, and measured service. Therefore, the main ad-

vantages of cloud computing are the availability of

high computational power and large storage, paying

only for what is used. The elasticity has a funda-

mental function in this context, allowing the growth

or the decrease of resources dynamically. But cloud

From the Sky to the Ground: Comparing Fog Computing with Related Distributed Paradigms

159

computing also has some constraints. The fundamen-

tal limitation is the connectivity between the cloud

and the end devices and users, because public cloud

providers are supported by large data centers around

the world, but not close enough to the end-user, result-

ing in Quality-of-Service (QoS) degradation, which

is not well-suited for time-critical service requests

(Mukherjee et al., 2018).

Another relevant aspect is that the cloud comput-

ing paradigm is a centralized computing model, and

most of the computations happen in the cloud. Al-

though data processing speeds have risen rapidly, net-

work bandwidth, even considering a high speed con-

nection like 5G, is becoming the bottleneck of cloud

computing for such a huge amount of data (Hu et al.,

2017). Overcoming these limitations will beneﬁt par-

ticular use cases as connected vehicles, smart cities,

virtual reality and healthcare (Naha et al., 2018). The

cloud computing characterization is showed in Figure

Figure 2: Cloud Computing Characterization.

4 FOG COMPUTING

Fog is a cloud close to the ground. Similarly the term

fog computing refer to a computation that operates at

the edge of the network, bringing cloud-like services

to be executed close to the end-users. Furthermore,

it provides computing resources for applications that

cannot perform properly with the high latency pro-

vided by cloud-only environments (Naha et al., 2018).

Bonomi et al. (Bonomi et al., 2012) presented the

ﬁrst deﬁnition of fog computing saying that it is a

highly virtualized platform that provides computing,

storage, and networking services among many com-

puting data centers or end-devices.

Various researchers have expanded and revised

this initial deﬁnition of fog computing. Vaquero et

al. (Vaquero and Rodero-Merino, 2014) and Yi et al.

(Yi et al., 2015) consider fog computing as a scenario,

composed of a high number of decentralized and het-

erogeneous devices, where they communicate and co-

operate among themselves and with the network to

perform data processing and storage without third-

party interventions. Services, applications, or basic

network functions that run in a sandboxed environ-

ment, can be supported by the data processing and

storage.

In the deﬁnition presented by Dastjerdi et al.

(Dastjerdi et al., 2016), fog computing is considered

a distributed computing paradigm. In this paradigm

the services provided by the cloud are essentially ex-

tended to the network edge. Fog computing addresses

application requirements that need low latency with a

huge and dense geographical distribution. Due to this,

fog computing supports computing resources, differ-

ent communication protocols, mobility, interface het-

erogeneity, integration with the cloud, and distributed

data analytics.

For Naha et al. (Naha et al., 2018), fog comput-

ing is a distributed platform where the edge or end

devices, that can be virtualized or not, will do most

of the processing. It resides between the cloud and

users and the cloud will do long-term storage and non-

latency-dependent processing.

Finally, fog computing was deﬁned in NIST (Iorga

et al., 2018) as a layered model facilitating the deploy-

ment of applications and services that are latency-

aware and distributed. This model enables ubiqui-

tous access to shared devices of scalable computing

resources that are not perceptibly different from each

other, although the extremes are quite distinct. There-

fore, fog computing provides, for the end-devices,

local computing resources and network connectivity

to centralized services, when needed, minimizing the

request-response time.

In all deﬁnitions, it is possible to note that fog

computing is tightly linked to the existence of a cloud

since fog can never replace the cloud completely as

we still need it to handle big or complex data prob-

lems (Gill et al., 2019). So, it is possible to state that

fog computing is suitable to be used when the cloud

does not meet the time limit, bandwidth limitations,

or latency requirements.

Considering all concepts presented, the following

fog computing deﬁnition is proposed: Fog computing

is a distributed architecture that uses the computa-

tional resources of devices located between end-users

and the cloud to optimize the processing and to re-

duce applications’ response times, meeting demands

that until now were not possible.

4.1 Fog Architecture

According to the NIST deﬁnition (Iorga et al., 2018),

one fundamental component in fog architecture is the

fog node. A fog node is any hardware device in a

fog computing environment that has system and hard-

ware resources added to high communication capabil-

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

160

ity (Bachiega et al., 2021).

The fog environment can be based on a software-

deﬁned fog architecture and use Fog Radio Access

Networks (F-RAN) (Mukherjee et al., 2018). How-

ever, layered (or hierarchical) representation is con-

sidered by Naha et al. (Naha et al., 2018) a bet-

ter way to represent fog architecture. Although this

topic has been largely researched in academia (Dast-

jerdi et al., 2016) and a three-tier architecture has been

commonly used to represent a fog system (Mahmud

and Buyya, 2016), it is possible to ﬁnd proposals with

four (Tang et al., 2015), ﬁve (Naas et al., 2017) or,

even, six (Fan et al., 2018) layers, as presented in Fig-

ure 3. A comprehensive review of fog computing ar-

chitectures can be found in (Habibi et al., 2020).

Figure 3 demonstrates that although there are vari-

ations in the number of layers in the existing architec-

tures, it occurs only in the Fog Layer, with some com-

ponents being more diluted and others being more

concatenated. Another important point is that the

End-users and Cloud layers are present in all archi-

tectures. Thereby, regardless of the number of lay-

ers being proposed, the architectures can be summa-

rized in three essential layers: End-users, Fog, and

Cloud. This three-layer architecture model will be

used throughout this article to allow comparison with

other computational paradigms.

The IoT/End-users Layer, at the base, represents

all IoT devices. It is in this layer that end-users re-

quest services that will be processed in the Fog and

the Cloud Layers. The Fog Layer acts as a link be-

tween IoT/End-users and the Cloud layers to provide

the necessary extra functionalities for application-

speciﬁc processing, as ﬁltering and aggregation be-

fore transferring the data to the cloud (Al-Doghman

et al., 2016). It is composed of fog nodes and this

layer comprises ‘intelligent’ devices that are capable

of computing, processing, and storing data in addi-

tion to routing and forwarding the data packets to the

upper tier (Sarkar and Misra, 2016).

Finally, the Cloud Layer has powerful computa-

tional resources to process all requests and responses

made by and sent to the Fog and IoT/End-users Lay-

ers directly. The Cloud Layer is a requirement in a

fog system. Similar paradigms that enable application

provisioning at the edge, as cloudlet (Satyanarayanan

et al., 2009) or Mobile Edge Computing (MEC) (Beck

et al., 2014) can operate in standalone mode.

4.2 Fog Characteristics

Bonomi et al. (Bonomi et al., 2012), writing for the

ﬁrst time about the fog computing paradigm, char-

acterized it into ten items, namely location aware-

ness and low latency; geographical distribution; large-

scale sensor networks; large number of nodes; support

for mobility; real-time interactions; predominance of

wireless access; heterogeneity; interoperability and

federation; and support for on-line analytics and in-

terplay with the cloud.

Since then, the number of studies about fog com-

puting has grown and revised these characteristics

(Hu et al., 2017). Recently, NIST (Iorga et al.,

2018) deﬁned the following six characteristics that are

considered essential in distinguishing fog computing

from other computing paradigms:

• Low latency: offering the lowest possible latency

due to the awareness of logical location and la-

tency costs for communication between fog nodes

in the context of the entire system;

• Geographical distribution: fog computing de-

mands widely distributed deployments for ser-

vices and applications;

• Heterogeneity: processing data acquired in differ-

ent forms through multiple types of network com-

munication capabilities;

• Interoperability: seamless support of certain

services requires the cooperation of different

providers;

• Real-time interactions: applications of fog com-

puting involve real-time interactions rather than

batch processing;

• Scalability: must be adaptive and support elastic

computing, resource pooling, data-load changes,

and network condition variations.

The fog computing conceptual model, proposed

by NIST (Iorga et al., 2018), also deﬁnes the pre-

dominance of wireless access and support for mobil-

ity as additional characteristics often associated with

fog computing. A fog computing characterization

is showed in Figure 4. Considering these charac-

teristics, some use-cases for fog computing include

smart cars, trafﬁc control, smart cities, smart build-

ings, surveillance and security. Furthermore, it is im-

portant to note that the sharp growth of 5G technology

will leverage adoption of IoT-related services and fog

computing is very appropriate to this scenario (Santos

et al., 2018).

5 RELATED PARADIGMS AND

TECHNOLOGIES

Apart from fog computing, there are other paradigms

based on a close proximity to connected devices and

From the Sky to the Ground: Comparing Fog Computing with Related Distributed Paradigms

161

Figure 3: Variation in the number of layers in Fog Computing Architectures.

Figure 4: Fog Computing Characterization.

this causes confusion about their concepts and deﬁ-

nitions. To avoid this confusion, next sections will

present details about each related paradigm and com-

pare them with the fog computing characteristics, as

well as place them in the three-tier architecture pre-

sented in section 4.1 to clarify understanding.

5.1 Edge Computing

Fog computing is often erroneously confused with

edge computing, but there are key differences be-

tween the two concepts. While fog computing runs

applications in a multi-layer architecture, edge com-

puting runs speciﬁc applications in a ﬁxed location,

that is, in the edge devices.

It can also be considered that the edge computing

can be limited to a few number of end-user devices

while fog computing has a bigger number of periph-

eral devices with hierarchical architecture. Therefore,

it is possible to see that edge computing is more re-

stricted than fog computing.

In edge computing the devices produce and con-

sume data, participating in the processing. Data stor-

age, computation ofﬂoading, processing, and caching

will be done by an edge node. The edge device is also

capable of distributing requests and providing cloud

services to the users (Avasalcai et al., 2020).

Furthermore, in edge computing, the devices are

not able to implement multiple IoT applications, since

there are limited resources and this can result in re-

source contention and increase processing latency.

On the other hand, fog computing can overcome these

limitations by seamlessly integrating edge devices

and cloud resources (Dastjerdi et al., 2016).

Finally, edge computing focuses on the end-

devices level, while fog computing focuses on the in-

frastructure level (Shi et al., 2016). The edge comput-

ing characterization is presented in Figure 5.

Figure 5: Edge Computing Characterization.

5.2 Multi-access Edge Computing

(MEC)

Previously, also referenced by some authors as “Mo-

bile Edge Computing”, MEC can be deﬁned as an

implementation of edge computing to bring compu-

tational and storage capacities to the edge of the net-

work within the Radio Access Network (RAN) to re-

duce latency and improve context-awareness (Dolui

and Datta, 2017).

According to Beck et al. (Beck et al., 2014), MEC

is a proposal for co-locating, at the base stations of

cellular networks, computing, and storage resources.

MEC is the evolution of mobile base stations and col-

laborative deployment of telecommunication and IT

networking. They operate on the edge of the Internet

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

162

and remain functional even without Internet connec-

tivity (Dolui and Datta, 2017). Figure 6 shows the

MEC characterization.

The MEC computing paradigm can provide sev-

eral services including IoT, location services, aug-

mented reality, caching service, video analytics, and

local content distribution. It can also provide low-

latency access in real-time to local content or by

caching the content at the MEC server. However,

some restrictions cannot be ignored, such as the in-

stallation of the MEC server, which is speciﬁcally

dedicated to them. The increase in resource demand

over time also becomes a major scaling problem (Cui

et al., 2021).

Figure 6: Mobile Edge Computing Characterization.

5.3 Mobile Cloud Computing (MCC)

The mobile computing concept proposed by Satya-

narayanan in 1996 (Satyanarayanan, 1996) represents

the computation performed via mobile, portable de-

vices, such as laptops, tablets, or mobile phones.

However, the evolving requirements of connected

devices make mobile computing alone not enough

to address some computing challenges of our days

(Yousefpour et al., 2019).

Mobile computing has gained a valuable comple-

ment as cloud computing matured. This combination

resulted in MCC, which is deﬁned as an infrastruc-

ture where both data storage and data processing oc-

cur outside of the mobile device (Habibi et al., 2020).

The cloud computing, the mobile computing, and

the wireless communication are combined to provide

the MCC (Mahmud and Buyya, 2016), improving the

Quality of Experience (QoE) of mobile users.

Mobile computing requires changes to some char-

acteristics of cloud computing, like the existence of

a low-latency intermediary tier, the optimization of

cloud infrastructure for running mobile application,

and the seamless ofﬂoading and remote execution.

The viability of MCC is based on a reliable end-to-

end network with high bandwidth and this can be

achieved by using virtual machines and cloudlets that

must be located closer to the mobile devices (Satya-

narayanan, 1996). Figure 7 presents a characteriza-

tion of MCC.

Figure 7: Mobile Cloud Computing Characterization.

5.4 Mobile Ad hoc Cloud Computing

(MACC)

MCC has a pervasive nature, but there are scenarios

where it is not suitable, for example where there is no

centralized cloud, or the infrastructure is insufﬁcient

(Yousefpour et al., 2019). An Ad hoc mobile network

consists of nodes that form a temporary, dynamic net-

work through routing and transport protocols, build-

ing a decentralized form of network (Hubaux et al.,

2001).

Therefore, MACC consists of a pool of devices

with high computational capabilities that are closer to

the user. This low-cost computational environment

is deployed over a network where all nodes cooper-

atively maintain the network (Balasubramanian and

Karmouch, 2017).

MACC’s motivation is to address situations in

MCC for which connectivity to cloud environments is

not feasible, such as the absence of or an intermittent

network connection (Yaqoob et al., 2016). Further-

more, the hardware used, the service access method,

and the distance from users are also other differences

between MACC and MCC. MACC’s characterization

is presented in Figure 8.

Finally, when compared with the fog computing,

connected devices in MACC are more decentralized,

and this allows the devices to form a more dynamic

network in places of sparsely connected devices or a

constantly changing network (Hubaux et al., 2001).

Figure 8: Mobile Ad hoc Cloud Computing Characteriza-

tion.

5.5 Mist Computing

Mist computing can be deﬁned as a lightweight and

rudimentary form of fog computing (Ranaweera et al.,

From the Sky to the Ground: Comparing Fog Computing with Related Distributed Paradigms

163

2021). It brings the fog computing layer closer to the

smart end-devices, using microcomputers and micro-

controllers to get it, beyond to reside directly within

the network at the edge of the network (Iorga et al.,

2018).

In the words of Yousefpour et al. (Yousefpour

et al., 2019), Mist computing can be referenced as

“IoT computing” or “things computing” and it could

be seen as the ﬁrst location where the computation

takes place in the IoT-fog-cloud continuum.

The data transfer uses a lot of battery power and

it is desirable that data can be processed, precondi-

tioned, and optimized ﬁrst before being stored. This

fact results in a data transfer much smaller, consum-

ing less power (Silva et al., 2017). The mist com-

puting paradigm can increase the autonomy of solu-

tions, and further decrease the latency period (Preden

et al., 2015). The mist computing characterization is

showed in Figure 9.

Figure 9: Mist Computing Characterization.

5.6 Cloudlet Computing

In 2009, Carnegie Mellon University introduced the

“data center in a box” concept, named Cloudlet Com-

puting (Satyanarayanan et al., 2009). Cloudlets are

composed of decentralized and widely dispersed In-

ternet infrastructure. Nearby mobile computers can

leverage their compute cycles and storage resources.

Cloudlets offer low latency for applications because

they are provided over one-hop access with high

bandwidth (Alli and Alam, 2020).

The mobile device, acting as a thin client, can of-

ﬂoad computational tasks through a wireless network

to a cloudlet, deployed one hop away (Riaz et al.,

2019). However, a cloudlet’s presence in mobile de-

vices’ proximity is necessary, as the end-to-end re-

sponse time, when executing applications, must be

smaller and predictable. If a device goes out of range

of a cloudlet, then it should gracefully switch to the

distant cloud, or worst-case scenario, rely solely on

its resources (Bilal et al., 2018).

Another important characteristic of cloudlets is

that they support local services for mobile clients by

dividing tasks among cloudlet nodes near the mobile

devices (Li and Wang, 2014). However, fog comput-

ing offers a more generic alternative that natively sup-

ports large amounts of trafﬁc and allows resources to

be anywhere along the thing-to-cloud continuum (Cui

et al., 2017). Cloudlet characterization can be seen in

Figure 10.

Figure 10: Cloudlet Computing Characterization.

5.7 Dew Computing

In the cloud computing spectrum, the on-premises

computer software-hardware organization paradigm

is known as Dew computing (Ray, 2017). In dew

computing, the on-premises computational resources

provide functionality which is independent, i.e. that

works seamlessly without an internet connection, and

also collaborative with cloud services, e.g. when an

internet connection is available it can synchronize

data and update a copy in the cloud.

So, dew computing intends to fully realize the po-

tentials of on-premises computational resources and

cloud services (Wang, 2016) acting together without a

permanent connection. The nature of dew computing

applications can be precisely described by the inde-

pendence that indicates their applications are inher-

ently distributed and the collaboration that indicates

that the applications are inherently connected.

Dew computing takes the concepts of service,

storage, and network, and goes further, deﬁning a

sub-platform, based on micro-services and distribut-

ing vertically its computing hierarchy (Wang, 2015).

The dew computing paradigm makes the use of hard-

ware resources easier since they are connected to a

network, covering a broad range of ad-hoc-based net-

working technologies (Skala et al., 2015). Figure 11

shows the characterization of dew computing.

Figure 11: Dew Computing Characterization.

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

164

6 FOG COMPUTING AND

RELATED PARADIGMS

COMPARISON

Based on the main fog characteristics presented

in Section 4.2, a comparison of aforementioned

paradigms is presented in Table 1. The idea is to

highlight the differences and similarities between the

paradigms and help clarify the way they compare to

fog computing.

Table 1: Fog computing and related paradigms comparison.

Paradigms

Low Latency

Geo-Distribution

Heterogeneity

Interoperability

Real-time

Scalability

Sky - - X X - X

Cloud - - X X - X

Fog X X X X X X

Edge X X X X X X

MEC X X - X X -

MCC - - X X - -

MACC - X - X - -

Mist - X X - X -

Cloudlet X X - X X X

Dew - - - X X -

Analyzing Table 1, it is possible to notice that

edge computing is the paradigm closer to fog com-

puting. On the opposite side, MCC, MACC and dew

computing share few characteristics of fog comput-

ing, moving away from the scope of this paradigm.

Finally, Figure 12 shows the position of each an-

alyzed paradigm on the IoT-Cloud continuum. Fog

computing is located between cloud and IoT/end-

users and, although the other described paradigms are

suitable for some speciﬁc use cases, fog computing

has been seen as a more general form of computing

due to its comprehensive deﬁnition scope (Yousef-

pour et al., 2019).

7 CHALLENGES

Fog is a distributed paradigm that extends cloud

computing to the network edge, providing services

closer to end users. But the end users can be dis-

tributed and connected to the network by different

ways and means, according to the use case they are

executing: health monitoring use cases, industrial IoT

(IIoT), Smart Power Grid, Smart Cities among oth-

ers (Avasalcai et al., 2020). Each use case can have

speciﬁc requirements and all of them together cre-

ate a range of options and needs that must be con-

sidered on the same fog environment, or on a feder-

ation of fog environments. In such a scenario, there

could be no need to consider mobility and the con-

nection to the fog can be considered as reliable. A

much more different scenario is a vehicular applica-

tion connected to a fog node located as a Road Side

Unit (RSU), where mobility and wireless communi-

cation are fundamental characteristics. Although the

communication architecture can be the same between

these two use cases, in a 3-tier architecture as pre-

sented on Section 4.1, the functionalities’ set needed

by each of them, as well as the options for privacy, se-

curity, fault tolerance, for instance, are different and

must be provided by the fog computing environment

accordingly.

A relevant challenge of fog computing is related to

security and privacy warranty. Heterogeneity of user-

equipment and access network makes it difﬁcult and

complex to implement security and privacy defense

features (Ranaweera et al., 2021). Moreover, once

the features are deployed, there is a trade-off between

the functionalities running on the user device and en-

ergy consumption. For example, the stronger the key

and the encryption process used, more energy from

user’s device will be consumed. Also, the fewer re-

sources the device has available to run these security

processes, the longer it takes to run them, increasing

the response time of an application or service being

executed and causing a worse QoE (Mahmud et al.,

2018). According to privacy and security solutions

in place, a minimum resource speciﬁcation must be

guaranteed by the service orchestrator when allocat-

ing a fog node to run a service.

Low latency is a requirement to run real time ser-

vices on fog infrastructures. To meet this require-

ment, the service is placed closer to the user and,

in case of user mobility, e.g., a mobile phone inside

a moving car, the service could be migrated to an-

other fog node that is located closer to the user’s new

place from time to time. Service migration is also

used as a strategy of ofﬂoading services to a resource

richer node and in case of node failure. The chal-

lenge is to guarantee the SLA/QoE while providing

a secure communication channel between fog nodes

(Ranaweera et al., 2021).

Finally, from the resource’s management perspec-

tive, a wide-open issue must be improved. Although

the virtual machine concept is extensively in use on

cloud computing architectures and still being used in

fog environments, recent works indicate that the use

of containers, a standard unit of software that pack-

ages up code and all its dependencies (Yin et al.,

From the Sky to the Ground: Comparing Fog Computing with Related Distributed Paradigms

165

Figure 12: Position of computing paradigms on IoT-Cloud continuum.

2018), is more appropriate for fog features and needs.

The use of federation, a set of public and private

providers connected through the Internet, is widely

adopted on cloud computing (Rosa et al., 2018). The

development of a federation for fog computing is

needed to ensure a greater computational capacity for

this technology. Scalability may also depend on such

a federation. Fog nodes help IoT devices lessening

their load and providing fast response. This allows

a higher number of devices to connect to the envi-

ronment without overloading cloud network. But the

challenge is to recognize when the number of de-

vices connected directly to a fog node has reached

a threshold in which may overload the node itself

(Fersi, 2021), therefore requests must be redirected

to other available fog nodes, including the federated

ones. Automatic service orchestration and resource

management are also needed (Costa et al., 2022).

8 RELATED WORK

In recent years, an increasing number of studies have

been done about fog computing and related paradigms

and technologies. Most of them, as in (Javed and

Mahmood, 2021) and (Fersi, 2021), comparing these

paradigms to cloud computing.

In Hu et al. (Hu et al., 2017), a study about fog

computing presents a comparison between fog, edge,

and cloud computing concepts. Some contextualiza-

tion about the fog architecture, characteristics, and

applications is presented.

Yousefpour et al. (Yousefpour et al., 2019) present

similarities and differences between fog computing

and the following paradigms: cloud computing, mo-

bile computing, edge computing, mobile cloud com-

puting, multi-access edge computing, mobile ad hoc

cloud computing, and mist computing.

A comparison between fog computing, multi-

access edge computing, and cloudlet is presented

by Mouradian et al. (Mouradian et al., 2018) and

(Ranaweera et al., 2021). In Bilal et al. (Bilal et al.,

2018) the micro data center concept is compared to

fog computing. In Naha et al. (Naha et al., 2018) the

fog concept is compared with dew computing. Fog,

edge, and mist Computing are compared in (Alli and

Alam, 2020).

Gill et al. (Gill et al., 2019) presents three emerg-

ing paradigms, blockchain, IoT, and artiﬁcial intelli-

gence, exploring how they will transform cloud com-

puting to solve complex problems of next-generation

computing. A complete study of related paradigms

was done by the authors, including fog computing.

Habbib et al. (Habibi et al., 2020) present a com-

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

166

Table 2: Related work.

Paper

Sky

Cloud

Fog

Edge

MEC

MCC

MACC

Mist

Cloudlet

Dew

(Hu et al., 2017) X X X

(Mouradian et al., 2018) X X X

(Bilal et al., 2018) X X X

(Naha et al., 2018) X X

(Yousefpour et al., 2019) X X X X X X X

(Gill et al., 2019) X

(Habibi et al., 2020) X X X

(Alli and Alam, 2020) X X X X

(Javed and Mahmood, 2021) X X

(Ranaweera et al., 2021) X X X X

(Fersi, 2021) X X

(Ogundoyin and Kamil, 2021) X X X X X X X X X

This paper X X X X X X X X X X

prehensive survey addressing the fog architectural

concepts, reviewing distinctions between cloud, edge,

mobile edge, and fog, and proposing a taxonomy for

architectural, algorithmic, and technological aspects

of fog computing. An evaluation of the current state

of the technology and identiﬁcation of the potential

future direction was also presented.

Finally, the article (Ogundoyin and Kamil, 2021)

comments on the possible methods for optimizing fog

computing and brings its applications, making met-

rics of its research in the literature and testing differ-

ent algorithms for solutions in various environments

and bringing with it proposals for future trends for

optimization of computing in fog.

Table 2 summarizes the computational paradigms

that were compared in each related work presented in

this section, ordered considering the year of each pub-

lication. All papers analyzed in this section compare

the fog computing concept with few other paradigms.

In contrast, this paper shows a comparison between

fog and all other relevant related paradigms, present-

ing the main differences and similarities. Further-

more, this paper is the only one that brings the con-

cepts of the Sky Computing.

9 CONCLUSIONS

The distance of cloud computing data centers from

the end-user’s devices prompt the increase of com-

putational technologies that solve this issue. In this

paper, we present concepts, similarities, and differ-

ences of nine computational paradigms and technolo-

gies that ﬁll in the gap between cloud computing and

the edges.

Among the compared technologies in this paper,

fog computing proved to be one of the most promis-

ing to be used closer to end-users, complementing the

services offered by cloud computing and also by sky

computing. It is a decentralized computing infrastruc-

ture that considers the best place between data source

and cloud to distribute storage, computation, and ap-

plications.

Furthermore, the concatenated three-tier fog com-

puting architecture with the related paradigms pre-

sented in Figure 12, helps understanding the posi-

tion and characterization of each analyzed paradigm.

Just like any emerging technology, these computa-

tional paradigms have some challenges to overcome,

as standardization, security, resource management,

QoS, and fault tolerance.

Unlike other related computational paradigms,

fog computing is both complementary to and an ex-

tension of traditional cloud-based models, being suit-

able to support many use-cases, mainly accelerated by

the increase of the 5G technology with IoT perspec-

tives.

Just as it was done for sky computing paradigm,

keeping up with the development of emergent com-

putational paradigms and new technologies may be

considered for future works to overcome the evolving

of the computation challenges.

REFERENCES

Al-Doghman, F., Chaczko, Z., Ajayan, A. R., and

Klempous, R. (2016). A review on fog comput-

ing technology. In 2016 IEEE International Confer-

ence on Systems, Man, and Cybernetics (SMC), pages

001525–001530. IEEE.

Alli, A. A. and Alam, M. M. (2020). The fog cloud of

things: A survey on concepts, architecture, standards,

tools, and applications. Internet of Things, 9:100177.

Avasalcai, C., Murturi, I., and Dustdar, S. (2020). Edge and

From the Sky to the Ground: Comparing Fog Computing with Related Distributed Paradigms

167

fog: A survey, use cases, and future challenges. Fog

Computing: Theory and Practice, pages 43–65.

Bachiega, J., da Costa, B. G. S., and Araujo, A. P. F. (2021).

Computational perspective of the fog node. 22nd In-

ternational Conference on Internet Computing & IoT.

Balasubramanian, V. and Karmouch, A. (2017). An infras-

tructure as a service for mobile ad-hoc cloud. In IEEE

7th Annual Computing and Communication Workshop

and Conference (CCWC), pages 1–7.

Beck, M. T., Werner, M., Feld, S., and Schimper, S. (2014).

Mobile edge computing: A taxonomy. In Proc. of the

Sixth International Conference on Advances in Future

Internet, pages 48–55. Citeseer.

Bilal, K., Khalid, O., Erbad, A., and Khan, S. U. (2018).

Potentials, trends, and prospects in edge technologies:

Fog, cloudlet, mobile edge, and micro data centers.

Computer Networks, 130(2018):94–120.

Bonomi, F., Milito, R., Zhu, J., and Addepalli, S. (2012).

Fog computing and its role in the internet of things. In

Proceedings of the First Edition of the MCC Workshop

on Mobile Cloud Computing, MCC ’12, pages 13–16.

Cloudify (2021). Getting started.

Costa, B., Bachiega, J., de Carvalho, L. R., and Araujo, A.

P. F. (2022). Orchestration in fog computing: A com-

prehensive survey. ACM Comput. Surv., 55(2).

Cui, M., Fei, Y., and Liu, Y. (2021). A survey on secure de-

ployment of mobile services in edge computing. Se-

curity and Communication Networks, 2021.

Cui, Y., Song, J., Ren, K., Li, M., Li, Z., Ren, Q., and

Zhang, Y. (2017). Software deﬁned cooperative of-

ﬂoading for mobile cloudlets. IEEE/ACM Transac-

tions on Networking (TON), 25(3):1746–1760.

Dastjerdi, A. V., Gupta, H., Calheiros, R. N., Ghosh, S. K.,

and Buyya, R. (2016). Fog Computing: Principles,

architectures, and applications. Internet of Things:

Principles and Paradigms, pages 61–75.

de Carvalho, L. R. and de Araujo, A. P. F. (2020). Perfor-

mance comparison of terraform and cloudify as mul-

ticloud orchestrators. In 2020 20th IEEE/ACM Inter-

national Symposium on Cluster, Cloud and Internet

Computing (CCGRID), pages 380–389. IEEE.

Dolui, K. and Datta, S. K. (2017). Comparison of edge

computing implementations: Fog computing, cloudlet

and mobile edge computing. GIoTS 2017 - Global

Internet of Things Summit, Proceedings.

Elkhatib, Y. (2016). Deﬁning cross-cloud systems. CoRR,

abs/1602.02698.

Fan, Y., Zhu, Q., and Liu, Y. (2018). Cloud/fog computing

system architecture and key technologies for south-

north water transfer project safety. Wireless Commu-

nications and Mobile Computing, 2018.

Fersi, G. (2021). Fog computing and internet of things in

one building block: a survey and an overview of inter-

acting technologies. Cluster Computing, pages 1–31.

Foster, I., Zhao, Y., Raicu, I., and Lu, S. (2008). Cloud

computing and grid computing 360-degree compared.

arXiv preprint arXiv:0901.0131.

Gill, S. S., Tuli, S., Xu, M., Singh, I., Singh, K. V., Lindsay,

D., Tuli, S., Smirnova, D., Singh, M., Jain, U., et al.

(2019). Transformative effects of iot, blockchain and

artiﬁcial intelligence on cloud computing: Evolution,

vision, trends and open challenges. Internet of Things,

8:100118.

Grozev, N. and Buyya, R. (2014). Inter-cloud architectures

and application brokering: taxonomy and survey. Soft-

ware: Practice and Experience, 44(3):369–390.

Habibi, P., Farhoudi, M., Kazemian, S., Khorsandi, S., and

Leon-garcia, A. (2020). Fog Computing : A Compre-

hensive Architectural Survey. IEEE Access, PP:1.

Hu, P., Dhelim, S., Ning, H., and Qiu, T. (2017). Survey

on fog computing: architecture, key technologies, ap-

plications and open issues. Journal of Network and

Computer Applications, 98(April):27–42.

Hubaux, J.-P., Gross, T., Le Boudec, J.-Y., and Vetterli,

M. (2001). Toward self-organized mobile ad hoc net-

works: the terminodes project. IEEE Communications

Magazine, 39(1):118–124.

Iorga, M., Feldman, L., Barton, R., Martin, M., Goren,

N., and Mahmoudi, C. (2018). The nist deﬁnition of

fog computing. Technical report, National Institute of

Standards and Technology.

Javed, Z. and Mahmood, W. (2021). A survey based study

on fog computing awareness. International Jour-

nal of Information Technology and Computer Science

(IJITCS), 13(2):49–62.

Keahey, K., Tsugawa, M., Matsunaga, A., and Fortes, J.

A. B. (2009). Sky computing. In IEEE Internet Com-

puting, page 43–51. IEEE Computer Society.

Kritikos, K. and Plexousakis, D. (2015). Multi-cloud appli-

cation design through cloud service composition. In

2015 IEEE 8th Int. Conf. on Cloud Computing, pages

686–693.

Kushida, K. E., Murray, J., and Zysman, J. (2015). Cloud

computing: From scarcity to abundance. Journal of

Industry, Competition and Trade, 15(1):5–19.

Li, Y. and Wang, W. (2014). Can mobile cloudlets sup-

port mobile applications? In IEEE INFOCOM

2014-IEEE Conference on Computer Communica-

tions, pages 1060–1068. IEEE.

Mahmud, M. and Buyya, R. (2016). Fog Computing: A

Taxonomy, Survey and Future Directions.

Mahmud, R., Kotagiri, R., and Buyya, R. (2018). Fog com-

puting: A taxonomy, survey and future directions. In

Internet of everything, pages 103–130. Springer.

Mell, P., Grance, T., et al. (2011). The NIST deﬁnition of

cloud computing.

Michelino, D., Leon, J. C., and Alvarez, L. F. (2013). Im-

plementation and testing of openstack heat.

Mouradian, C., Naboulsi, D., Yangui, S., Glitho, R. H.,

Morrow, M. J., and Polakos, P. A. (2018). A Com-

prehensive Survey on Fog Computing: State-of-the-

Art and Research Challenges. IEEE Communications

Surveys and Tutorials, 20(1):416–464.

Mukherjee, M., Shu, L., and Wang, D. (2018). Survey of

fog computing: Fundamental, network applications,

and research challenges. IEEE Communications Sur-

veys and Tutorials, 20(3):1826–1857.

Naas, M. I., Parvedy, P. R., Boukhobza, J., and Lemarc-

hand, L. (2017). IFogStor: An IoT Data Placement

CLOSER 2022 - 12th International Conference on Cloud Computing and Services Science

168

Strategy for Fog Infrastructure. Proceedings - 2017

IEEE 1st International Conference on Fog and Edge

Computing, ICFEC 2017, (May 2018):97–104.

Naha, R. K., Garg, S., Georgakopoulos, D., Jayaraman,

P. P., Gao, L., Xiang, Y., and Ranjan, R. (2018).

Fog computing: Survey of trends, architectures, re-

quirements, and research directions. IEEE Access,

6:47980–48009.

Ogundoyin, S. O. and Kamil, I. A. (2021). Optimization

techniques and applications in fog computing: An ex-

haustive survey. Swarm and Evolutionary Computa-

tion, 66:100937.

Paraiso, F., Haderer, N., Merle, P., Rouvoy, R., and Sein-

turier, L. (2012). A federated multi-cloud paas infras-

tructure. In 2012 IEEE Fifth Int. Conf. on Cloud Com-

puting, pages 392–399.

Preden, J. S., Tammem

ae, K., Jantsch, A., Leier, M.,

Riid, A., and Calis, E. (2015). The beneﬁts of self-

awareness and attention in fog and mist computing.

Computer, 48(7):37–45.

Ranaweera, P., Jurcut, A. D., and Liyanage, M. (2021).

Survey on multi-access edge computing security and

privacy. IEEE Communications Surveys & Tutorials,

23(2):1078–1124.

Ray, P. P. (2017). An introduction to dew computing: Deﬁ-

nition, concept and implications. IEEE Access, 6:723–

737.

Riaz, N., Qaisar, S., Ali, M., and Naeem, M. (2019).

Node selection and utility maximization for mobile

edge computing–driven iot. Transactions on Emerg-

ing Telecommunications Technologies, page e3704.

Rosa, M. J., Ara

ujo, A. P., and Mendes, F. L. (2018).

Cost and time prediction for efﬁcient execution of

bioinformatics workﬂows in federated cloud. In 2018

IEEE International Conference on Bioinformatics and

Biomedicine (BIBM), pages 1703–1710. IEEE.

Santos, J., Wauters, T., Volckaert, B., and De Turck, F.

(2018). Fog computing: Enabling the management

and orchestration of smart city applications in 5g net-

works. Entropy, 20(1):4.

Sarkar, S. and Misra, S. (2016). Theoretical modelling of

fog computing: a green computing paradigm to sup-

port iot applications. Iet Networks, 5(2):23–29.

Satyanarayanan, M. (1996). Fundamental challenges in mo-

bile computing. Proceedings of the Annual ACM Sym-

posium on Principles of Distributed Computing, pages

1–7.

Satyanarayanan, M., Bahl, V., Caceres, R., and Davies, N.

(2009). The case for vm-based cloudlets in mobile

computing. IEEE pervasive Computing.

Shi, W., Cao, J., Zhang, Q., Li, Y., and Xu, L. (2016). Edge

computing: Vision and challenges. IEEE internet of

things journal, 3(5):637–646.

Shirinkin, K. (2017). Getting Started with Terraform. Packt

Publishing Ltd.

Silva, P. M. P., Rodrigues, J., Silva, J., Martins, R., Lopes,

L., and Silva, F. (2017). Using edge-clouds to re-

duce load on traditional wiﬁ infrastructures and im-

prove quality of experience. In 2017 IEEE 1st In-

ternational Conference on Fog and Edge Computing

(ICFEC), pages 61–67. IEEE.

Skala, K., Davidovic, D., Afgan, E., Sovic, I., and So-

jat, Z. (2015). Scalable distributed computing hier-

archy: Cloud, fog and dew computing. Open Journal

of Cloud Computing (OJCC), 2(1):16–24.

Tanenbaum, A. S. and Van Steen, M. (2007). Distributed

systems: principles and paradigms. Prentice-Hall.

Tang, B., Chen, Z., Hefferman, G., Wei, T., He, H., and

Yang, Q. (2015). A Hierarchical Distributed Fog

Computing Architecture for Big Data Analysis in

Smart Cities. Proceedings of the ASE BigData & So-

cialInformatics 2015, (October):28.

Vaquero, L. M. and Rodero-Merino, L. (2014). Finding

your way in the fog: Towards a comprehensive deﬁni-

tion of fog computing. SIGCOMM Comput. Commun.

Rev., 44(5):27–32.

Wang, Y. (2015). Cloud-dew architecture. International

Journal of Cloud Computing, 4(3):199–210.

Wang, Y. (2016). Deﬁnition and categorization of dew com-

puting. Open Journal of Cloud Computing (OJCC),

3(1):1–7.

Yaqoob, I., Ahmed, E., Gani, A., Mokhtar, S., Imran, M.,

and Guizani, S. (2016). Mobile ad hoc cloud: A sur-

vey. Wireless Communications and Mobile Comput-

ing, 16(16):2572–2589.

Yi, S., Li, C., and Li, Q. (2015). A Survey of Fog Comput-

ing. Proceedings of the 2015 Workshop on Mobile Big

Data - Mobidata ’15, pages 37–42.

Yin, L., Luo, J., and Luo, H. (2018). Tasks scheduling and

resource allocation in fog computing based on con-

tainers for smart manufacturing. IEEE Transactions

on Industrial Informatics, 14(10):4712–4721.

Yousefpour, A., Fung, C., Nguyen, T., Kadiyala, K., Jalali,

F., Niakanlahiji, A., Kong, J., and Jue, J. P. (2019).

All one needs to know about fog computing and re-

lated edge computing paradigms: A complete survey.

Journal of Systems Architecture, (December 2018).

From the Sky to the Ground: Comparing Fog Computing with Related Distributed Paradigms

169