Eﬃcient Utilization of Energy using Fog and Cloud
based Integrated Environment in Smart Grid
Ayesha Anjum Butt1, Nadeem Javaid1,∗, Amjad Rehman2and Tanzila Saba3
1COMSATS University Islamabad, Islamabad 44000, Pakistan
2MIS Department COBA, Al Yamamah University, Riyadh 11512, Saudi Arabia
3Prince Sultan University, Riyadh 11512, Saudi Arabia
Abstract—Smart Grid (SG) gives the opportunity of
two way communication to consumers and utility. SG
balances and monitors the consumption of electricity
of the consumers. The concept of fog and cloud is
introduced in order to enhance SG eﬃciency. Cloud
gives computational and storage facilities to the users.
The delay and latency issues arise because of its remote
location. To resolve these issues, fog is introduced by
cisco in recent years. The performance of cloud and fog
is enhanced using a diﬀerent technique. In this paper,
we introduce fog and cloud based environment for
the eﬃcient utilization of energy. In this environment,
Micro Grid (MG) is also attached to the buildings
to make their consumption eﬃcient using renewable
resources. MG makes these buildings and societies
smarter using fog and cloud based environment in
SG. Fog executes those request or allocates resources
to those which require less computational power and
resources; otherwise, these requests are forwarded to
the cloud. We apply Crow Search Algorithm (CSA) and
server broker policies to calculate the Response Time
(RT), Processing Time (PT) and cost.
Index Terms—Fog, Cloud, Smart Grid, Smart Soci-
eties, Load Balancing, Server Broker Policies
The evolution of Internet of Things (IoT) brings the
emergence of smart societies such as smart homes, smart
buildings, Smart Grid (SG), etc. The SG facilitates two
way communication between consumers and utilities to
consume and monitor the use of electricity by digital
communication. The system performs following functions:
reduce electricity consumption and cost, maximize the
consumer comfort and reliability of energy supply chain
which are done by this system . The SG is introduced
to beat the limitations of the traditional grids. Smart
meters are also used by the SG to enhance its eﬃciency
and performance. The SG monitors and balances the
consumption of the buildings of residential, industrial and
commercial areas. Smart meter keeps track of the electric-
ity consumption of smart buildings and homes and manage
the peak hours. SG provides help to both consumers and
utility by reducing the consumption cost of consumer and
energy usage of grid in an eﬃcient manner.
To make SG smarter, the concept of cloud computing
is introduced. By using this concept, the consumers of
diﬀerent regions can get the beneﬁts from other regions of
SG, if required. Cloud environment consists of computing
environment, which consists of numerous Virtual Machines
(VMs) and Physical Machines (PMs) inside them. Cloud
provides only computational and storage resources to
those consumers, who are utilizing the resource of it.
Furthermore, cloud provides resources to these consumers
who are particular user of it. When some other consumers
requests for resources, the delay increases due to maximum
requests. To overcome these limitations of cloud, the con-
cept of fog is introduced by "cisco" in 2014. The fog act
as an intermediary layer between cloud and consumer. It
provides ease to consumers in terms of reducing latency,
delay and provides storage also. Due to its smaller size
than cloud, diﬀerent societies can purchase their own fog.
In the SG environment, the fog and cloud provide
reliability to consumers to enhance their performance. The
user can execute multiple tasks on cloud and fog VMs and
PMs. The total number of user requests process faster with
less latency using fog. In literature, diﬀerent balancing
algorithms, techniques and services broker policies are
used in order to enhance the performance of cloud and
fog which reduce their computational cost . With the
passage of time, the cloud and fog based environment
becomes popular in industrial societies. To make industries
smarter, cloud and fog based environment is introduced in
industries ,  These environments are used to store
the big data of the industries and fog provides faster
computing services with less latency or delay.
To overcome all aforementioned challenges, we have
introduced fog and cloud integrated environment. We also
use MG with our proposed architecture. In this work, we
have considered the requests of residential societies. If the
processing time of these requests needs high computational
resources, then we transfer these request to cloud. MG
handles and monitors the consumption of industries and
homes. We use three service broker policies that are Clos-
est Data Center(CDC), Optimize RT (ORT) and Recon-
ﬁgure Dynamically with Load (RDL). The load balancing
algorithm is used to make our environment reliable and
scalable by adding ﬁxed number of VMs and PMs inside
them. The load balancing Crow Search Algorithm (CSA) is
implemented in the scenario of residential side. To simulate
2019 International Conference on Computer and Information Sciences (2019 ICCIS) 978-1-5386-8125-1 c
the results, we have used cloud analyst simulator.
The rest of the paper is organized as follows: Section II
deﬁnes related work, Section III problem statement and
motivation. The proposed system model deﬁnes in section
IV, Section V deﬁnes the CSA Load Balancing Algorithm,
Section VI deﬁnes the simulations results and discussion.
The last Section VI deﬁnes the conclusion of this paper.
II. Related Work
One of the most valuable resources of the modern
era is energy. In previous studies, most of the research
and discussion is about smart buildings and their energy
consumption. The authors in  discuss the smart cities
and number of smart buildings in it. The basic goal
of their work is to minimize the energy. To fulﬁll their
demands and desires,  proposes a scheme, in which
various cloud DCs are connected to the centralized cloud
and interconnected to the grid. In this proposed scheme,
authors use renewable energy resources to minimize the
energy consumption of the grid of the smart city. For this
purpose, they use a technique named as "Novel multi-
tenant cloud based Nano grid for smart buildings". By
using this technique and proposed scheme, they achieve
scalability and performance of the system. However, with
all these respective features, they can enhance their exe-
cution time. In , Boroojeni et al. discuss the increasing
demand of clusters of MGs for connectivity to increase
their ﬂexibility and security. In cluster, each MG is un-
aware about status and ﬂow. The contribution of their
struggle is that they propose oblivious routing algorithm.
In their adopted approach, the power optimal ﬂow problem
is solved while managing congestion and mitigating power
losses. Their proposed or adopted methodology works for
both radial and non-radial network disregarding of scale
and topology of MGs in the cluster. To evaluate the
performance and check the eﬀectiveness of their proposed
and adopted methodology, they use Matlab simulations.
OPAL-RT real time digital simulation system. To check
the communication path between MG and cloud, they
perform OMNET++ simulations. These simulations are
run and checked practically by implementing on their
proposed system of MG and cloud environment. The main
features of their proposed system are that, system is
eﬃcient and interoperable in terms of performance and
exchange information. However, their proposed system is
designed only for one way communication.
In Fog and cloud computing, SG becomes eﬃcient and
intelligent with the help of emerging technology of IoT.
In , Cao et al. propose an eﬀective privacy preserving
scheme for electric load monitoring which can diﬀerentiate
closer nodes in SG. In this proposed scheme, the Factorial
Hidden Markov Model (FHMM) is introduced. The main
objective of their proposed scheme is that they realize
the privacy towards fog in SG. After implementing the
scheme for achieving the key features of their proposed
technology, they propose a model of cloud, fog and SG.
To measure the performance of their scheme and check
the privacy level of their implemented model, they used
NILMTK tool on the basis of REDD data set. Subsequent
to this, their work is compared with other two schemes,
i.e., Barbosa’s and Sankar’s. They also use noise as a
parameter in their work to switch the states and check the
accuracy of their implemented scheme. The interactivity
of their proposed scheme and performance of the proposed
model is relatively better than the previous schemes and
models. On the other hand, the computational cost of their
implemented model is optimal high. The Yaghmaee et al.
in  propose two tier cloud based DSM for the customers
to control their residential load with less power generation
and high storage eﬃciency as an energy source. They
consider a power system which consists of multiple regions
and equipped with a number of MGs. According to their
research, edge cloud is utilized to ﬁnd the optimal power
in each region to schedule the customer appliances. In ,
they also propose two level optimization algorithm with a
multi level cost function. Their main aim is to minimize
the consumption cost of both consumer and utility. After
implementing the aforementioned proposed system model
and implemented algorithm, they improve the peak load
and peak to average ratio. To simulates the result of
their proposed algorithm, they use the Nash equilibrium
technique. On the other hand, they compromise the user
In , Djabir et al. implement a real time dynamic pric-
ing model for electric vehicles charging and discharging.
The purpose of this proposed model is to reduce the peak
load. Their implemented approach using decentralized
cloud computing architecture which is based on Software
Deﬁne Networking (SDN) and Network Function Virtu-
alization (NFV). Fog layer and SG are also part of their
proposed architecture. In their work, they aim to schedule
the requests between charging and discharging vehicles.
To fulﬁll their aims, they propose linear optimization
approach and real time pricing model. Stability is the main
feature of their work. On the other side, there is trade-oﬀ
between user comfort and cost.
A new identiﬁcation method for smart metering is pro-
posed in . This proposed method processes time and
numerical information for identiﬁcation of data. To de-
identiﬁed state for statistical processing and availability
of data, they use processing of database using standardize
query language. To further check the eﬃciency, interop-
erability and interactivity of their system, they propose
an architecture which consists of cloud and SG. The main
objective of their proposed method and architecture is re-
identiﬁcation of metering of smart data. After implemen-
tation of their proposed method, they achieve their ob-
jective, interactivity and interoperability of their system.
However, security is compromised in their system. In ,
Rekha et al. proposes a framework of cloud computing
in SG environment by creating small integrated energy
hub which supports to handle the storage of huge data.
They want to reduce the cost, save energy and handle
large data. For these purposes, they propose stochastic
dynamic programming. In their work, they also minimize
the cost and maximize the proﬁt for utility provider.
However, in their system, peak hours are ﬁxed and it
also compromises the user’s comfort. In the residential,
industrial and commercial domains energy management
is required to control their power generation and con-
sumption. According to them, the energy management
helps their system to reach zero net energy in residential
domain. The IoT becomes the revolutionary change in
cost and features using actuators and sensors. However,
scalability is still issues for them. Al Faruque et al. in-
troduce a novel platform for energy management in their
work. The implemented platform introduces the feature
of interoperability and scalability. To improve the market
cost and minimize the energy consumption in their work,
they introduce fog based architecture with MGs. Their
proposed architecture consists of cloud, fog and number
of MGs that are attached with residential and commercial
buildings. When these MGs are connected to buildings,
they control and monitor their electricity consumption.
In this work, they use home energy and MG energy
management system. By using the services of IoT, they
collect the data from commercial and residential buildings.
To simulate and calculate the energy consumption of data,
they use (Million Instruction Per Second) MIPS processor,
Raspberry Pi, Zig-Bee standard, etc, are used. However, to
calculate the energy consumption of homes and buildings
they consider one home instead of multiple homes. .
III. Motivation and Problem Statement
In  and - , proposed work discusses diﬀerent
challenges and problems that are related to smart facto-
ries, cloud based architecture and other.
The industries are converted into smart industries in
, . Cyber Physical System (CPS) interacts virtual
globe and tangible using predictive algorithm. The prob-
lems which are to be solved are equipment prognosis and
maintenance. However, there is another issue, how to solve
the problems over the internet. The range of activities
for product design is simulation, manufacturing, testing,
management and all other tasks in product life cycle .
However, no technique is proposed about enhancing and
assuring the production yield, which targets the product
life cycle. To solve the aforementioned challenges Ad-
vanced Manufacturing Cloud of Things (AMCOT) on the
basis of IoT, CPS and Cloud Computing is proposed in .
By adopting all these techniques the performance of the
industries is improved. Because of cloud platform, it han-
dles and stores large data. However, implementation cost
and delay to store large data are not discussed. To solve
the delay issue fog based integrated platform is introduced
in , . Because of its expertise and adaptable nature,
the number of incoming requests increases. So, it becomes
diﬃcult to schedule the tasks. To solve this issue PSO,
Energy Load Balancing (ELB), Energy Aware Dynamic
Task Scheduling (EDTS) and Critical Path Assignment
(CPA) are proposed in , ,  and . Still, cost,
performance, resource utilization are the major issues of
proposed algorithms and techniques.
This work is to devise a cloud and fog based integrated
environment using nature inspired algorithm CSA. The
aim of the proposed work is to overcome the issues of delay,
latency, system performance and minimization of cost in
terms of smart homes by utilizing the fog resources.
IV. Proposed System Model
In the proposed system model, we introduce cloud and
fog based integrated environment, which is depicted in Fig.
1. The cloud, fog, user and consumer layers are deﬁned in
this system model.
In this proposed architecture, to overcome the limita-
tions and improve the services of cloud, fog is introduced
and integrated with the cloud. This proposed architecture
describes the ﬂow of power in the smart city of Asia. This
architecture consists of three layers. Each layer deﬁnes
its own portrayal in the communication and power ﬂow
process. The ﬁrst layer of the architecture is known as
a cloud layer. This layer consists of the number of VMs
and PMs inside DCs. These DCs handles the requests of
the consumer that comes to the cloud from the bottom
two layers of this proposed model. This cloud provides the
communication services to fog and consumers. There is
also a two way communication between cloud and utility.
This utility act as a service provider and deals with the
requests of a cloud. There is two way communication
between cloud and utility. The cloud forwards the requests
to the consumer of SHs inside the Smart city of Asia.
The utility checks the demand of consumer according to
coming requests, then give information about the rates
of electricity that current time to the cloud. There is
also two way communication between cloud and fog. The
information about the rates of electricity forwards to the
fog from the cloud. According to the rates, the upcoming
requests to cloud again forwarded to the utility. The utility
provides a direct supply of the power to the required SH
in the particular block of the smart city. The cloud is on
the remote location of the consumer layer. The delay and
latency issue can occur during the ﬂow of requests from
consumer and information ﬂow from the cloud side.
To overcome the above mentioned issue, we introduce
number of fogs in our proposed system, it is because these
fogs depends on our deﬁned number of clusters in the
smart city. The second layer is known as a fog layer.
In this proposed architecture, there are the ﬁve numbers
of fogs at the edge of the network and provides services
according to the size of the requests with low latency. It
is because this layer is closest to the consumer layer and
it also improves the delay performance of the system. In
this proposed system, fog layer have the number of PMs
and there is a number of VMs in it, these PMs are number
Block 1 Block 2 Block 5
Cluster of Buildings
Cluster of Buildings
Smart City in Asia Region
Two Way Communication
Cluster of Buildings
Fig. 1: Cloud and Fog Based Integrated System Model
of CPUs which are used to process the consumer and user
data. There is a two way communication between fog and
cloud and fog is located at the edge of the network. So,
it processes the requests according to its capacity and
requests are referred to a cloud. The fog provides the given
information of related energy to the consumer of the SBs
attached with it. This proposed system also provides two
way communication between fog and clusters of buildings,
between MG and fog.
MG is attached to residential consumers and fog, it
converts renewable resources into useful energy resources
and gives it to the consumer.
The third layer of this proposed architecture, which is
known as the consumer layer. In this proposed architec-
ture, the consumer is known as a residential consumer.
There is two way communication between consumer and
fog. The consumer sends a request to the fog and gets
services according to the requests. This model consists
of ﬁve numbers of clusters and each cluster has a 600-
1000 number of requests. The consumer of this proposed
architecture sends requests to the fog. The fog processes
and executes the requests the requests which have high
demands send to the cloud and other fulﬁlled by the MG.
The blocks which are located inside this proposed ar-
chitecture holds all the three layers of this architecture.
The utility of this proposed architecture act as a service
provider. The power is sent to the SH located inside
the smart city. The outermost boundary of the proposed
architecture shows the overall design and working of a
smart city in the Asia region. The ﬂow of power is also
presented in this architecture that how utility supplies
power to the cluster of buildings in smart city.
V. CSA Load Balancing Algorithm
The novel meta heuristic algorithm that is based on
the nature of the crow is known as CSA. This algorithm
is proposed because the crow and corvids are considered
more intelligent birds. The algorithm works same as the
nature of the crows, i.e., crows search for food, steal and
hide the food from other crows. In this paper, CSA 1
performs load balancing. CSA initializes VMs, fogs in the
case of load balancing. Besides, the probability ﬁtness is
calculated with respect to DC. Our environment is fog
based, so the fogs act as a source of food. Therefore, the
implemented load balancing algorithm provides help to
VMs to ﬁnd the best feasible solution until they ﬁnd the
best ﬁtness value or best optimization solution value.
VI. Simulation Results and Discussion
In this paper, optimization algorithm is implemented to
utilize the energy resource eﬃciently. For this purpose we
implement CSA in this paper. In this section, we discuss
the RT, PT and cost on the basis of three policies. These
policies are CDC, ORT and RDL. To simulate the results
of our implemented algorithm region three is selected.
The reason to select the region three is that, this region
contains 27 million consumers. This region is known as
Asia. It has been further subdivides in several regions
that are central, South, East, Southern Asia, etc. These
continents contains large number of commercial, industrial
and residential smart buildings. The total population of
Algorithm 1: CSA
2Randomly initialize the position of a ﬂock of C
crows in the search space
3Search for list of VM and Fog
6for t = 1:24 do
7Let Yis a random position of search space
8Evaluate the position of the CVM
9Initialize the memory of each CVM
10 while iter ≤itermax do
11 Determine the VM_Size
12 Compute the processing time using equation
13 Calculate Transmission delay using equation
14 Compute the RT using equation 3
15 for i=1:Cdo
16 Randomly get a CVM jto follow i
17 Deﬁne awareness probability
18 if rj≥AP j,iter then
19 xi,iter1=xi,iter +riX(mj,iter −xi,iter)
21 xi,iter+1 =Y
25 Check the feasibility of new fog
26 Evaluate the new position of the CVM
this region is 43,810,582 who lives in the smart societies
of this region. To make the eﬃcient utilization of resources
and optimize energy diﬀerent algorithms are used.
In this paper, CSA is used for this purpose. On the
basis of this paper scenario, result is simulated on the basis
of ﬁve clusters and these cluster contains 600-1000 users.
After simulation on the basis of aforementioned policies,
it is proven that CSA has better RT and PT on the basis
of CDC and ORT policies. RDL has maximum RT and
PT in Fig. 2. The total cost of CDC and ORT is also
minimum as compared to RDL in Fig. 3. The RDL is not
fully implemented in cloud analyst scenario that’s why it
takes maximum RT, PT. Increase in cost has also same
issue as mention above.
The formula which is used to calculate the processing
time is given in below:
P T =TR /V Mspeed (1)
. Where TR is the total requests of the consumers and
V Mspeed is the speed of VMs. The transmission delay is
computed as given below:
T D =T otall+T otalT D (2)
Where lis latency and TD is the transmission delay.
The formula which is used to calculate the response time
RT =FT−AT+T D (3)
. Where F Tis the ﬁnish time, A Tis the arrival time and
TD is the transmission delay.
A. Response Time
In this section, the RT is calculated on the basis of
clusters. These clusters contain 600-1000 number of SHs.
The RT is calculated on the basis of number and size of
The Table I shows the avg. RT of clusters on the basis
of aforementioned three service broker policies. The RT of
CDC is 6.58%, RT of ORT is 7.72% and RT of RDL is
TABLE I: Avg. RT of clusters on the basis of Policies
User Base CDC (ms) ORT (ms) RDL (ms)
C1 79.60 93.59 1062.27
C2 79.63 93.63 1055.84
C3 79.78 94.24 1036.51
C4 80.01 93.58 1024.06
C5 79.95 93.50 1009.83
CDC ORT RDL
Service Broker Policies
Avg. RT and PT (ms)
Avg. RT (ms)
Avg. PT (ms)
Fig. 2: Average RT and PT
TABLE II: Avg. PT of fogs on the basis of Policies
Fogs CDC (ms) ORT (ms) RDL (ms)
Fog1 24.47 39.99 425.89
Fog2 31.53 45.07 1126.54
Fog3 31.89 45.38 1121.32
Fog4 31.64 45.43 1126.57
Fog5 31.39 45.39 1126.27
B. Processing Time
In this section, the PT of the DCs is deﬁned. The PT of
DCs depends on the number and size of the components.
The PT of DCs of all three policies; CDC, ORT and RDL
is shown in Table II. The PT of CDC is 2.84%, PT of ORT
is 4.17% and RDL is 92.97%. The RDL takes maximum
processing time as compared to other two policies because
ORT is not full implemented in cloud analyst tool.
In this section, the cost is deﬁned. This total cost
depends on the number of resources and data transfer cost,
etc. In Fig. 3, it is deﬁned that the total cost of policy
CDC and ORT is less than the cost of RDL. The VM
cost, Transfer cost and MG cost are also compared on the
basis of three policies in Fig. 3.
CDC ORT RDL
Service Broker Policies
Fig. 3: Total Cost
The industrial and commercial societies are becoming
smarter day by day. In this paper, a fog and cloud based
architecture is proposed. This architecture is used to make
an eﬃcient utilization of energy consumption in SG. MG
also plays an eﬃcient role in SG. We consider average RT,
PT and total cost of this proposed architecture. The CSA
is used to balance the load of a network along with three
service broker policies: CDC, ORT and RDL. The RT with
CDC is 66.04% and ORT is 77.53%. The PT with CDC is
2.84% and ORT is 4.15%. On the other hand RT and PT,
with CSA and RDL is not good. The average RT, PT and
total cost of RDL is 92%, 57.79% and 57.79%. In future,
we will consider VM migration problem.
 Stojkoska, Biljana Risteska, and Kire Trivodaliev. “A review of
Internet of Things for smart home: Challenges and solutions.”
Journal of Cleaner Production 140 (2017): 1454-1464.
 Zahoor, Saman, Sakeena Javaid, Nadeem Javaid, Mahmood
Ashraf, Farruh Ishmanov, and Muhammad Khalil Afzal. “Cloud
Fog Based Smart Grid Model for Eﬃcient Resource Manage-
ment.”Sustainability 10, no. 6 (2018): 1-21.
 Lin, Yu-Chuan, Min-Hsiung Hung, Hsien-Cheng Huang, Chao-
Chun Chen, Haw-Ching Yang, Yao-Sheng Hsieh, and Fan-
Tien Cheng. “Development of Advanced Manufacturing Cloud
of Things (AMCoT) A Smart Manufacturing Platform.” IEEE
Robotics and Automation Letters 2, no. 3 (2017): 1809-1816.
 Kapsalis, Andreas, Panagiotis Kasnesis, Iakovos Venieris, Dimi-
tra Kaklamani, and Charalampos Patrikakis. “A cooperative fog
approach for eﬀective workload balancing.” IEEE Cloud Comput-
ing 4, no. 2 (2017): 36-45.
 Kumar, Neeraj, Athanasios Vasilakos, and Joel JPC Rodrigues.
“A multi-tenant cloud-based DC nano grid for self-sustained
smart buildings in smart cities.” IEEE Communications Magazine
55, no. 3 (2017): 14-21.
 Boroojeni, Kianoosh, Hadi Amini, Arash Nejadpak, Tomislav
DragiÄŊeviÄĞ, Sundaraja Sitharama Iyengar, and Frede Blaab-
jerg. “A novel cloud-based platform for implementation of obliv-
ious power routing for clusters of microgrids.” IEEE Access 5
 Cao, Hui, Shubo Liu, Longfei Wu, Zhitao Guan, and Xiaojiang
Du. “Achieving diﬀerential privacy against non intrusive load
monitoring in smart grid: A fog computing approach.” Concur-
rency and Computation Practice and Experience (2018): e4528.
 Yaghmaee, Mohammad Hossein, Morteza Moghaddassian, and
Alberto Leon-Garcia. “Autonomous two-tier cloud-based demand
side management approach with microgrid.” IEEE Transactions
on Industrial Informatics 13, no. 3 (2017): 1109-1120.
 Chekired, Djabir Abdeldjalil, Lyes Khoukhi, and Hussein Mouf-
tah. “Decentralized cloud-SDN architecture in smart grid: A
dynamic pricing model.” IEEE Transactions on Industrial Infor-
matics 14, no. 3 (2018): 1220-1231.
 Lee, Donghyeok, Namje Park, Geonwoo Kim, and Seunghun
Jin. “De-identiﬁcation of metering data for smart grid personal
security in intelligent CCTV-based P2P cloud computing envi-
ronment.” Peer-to-Peer Networking and Applications (2018): 1-
 Reka, Sofana, and Ramesh. “Demand side management scheme
in smart grid with cloud computing approach using stochastic
dynamic programming.” Perspectives in Science 8 (2016): 169-
 Al Faruque, Mohammad Abdullah, and Korosh Vatanparvar.
“Energy management-as-a-service over fog computing platform.”
IEEE internet of things journal 3, no. 2 (2016): 161-169.
 Drath, and Horch. “Industrie 4.0: Hit or Hype?.” IEEE Indus-
trial Electronics Magazine, 8,(2014): 56-58.
 Hung, Min-Hsiung, Yu-Yung Li, Yu-Chuan Lin, Chun-Fan Wei,
Haw-Ching Yang, and Fan-Tien Cheng. “Development of a novel
cloud-based multi-tenant model creation service for automatic
virtual metrology.” Robotics and Computer-Integrated Manufac-
turing 44 (2017): 174-189.
 Xu, Xun. “From cloud computing to cloud manufacturing.”
Robotics and computer-integrated manufacturing 28, no. 1
 Wang, Shiyong, Jiafu Wan, Daqiang Zhang, Di Li, and Chunhua
Zhang. “Towards smart factory for industry 4.0: a self-organized
multi-agent system with big data based feedback and coordina-
tion.” Computer Networks 101 (2016): 158-168.
 Wan, Jiafu, Boxing Yin, Di Li, Antonio Celesti, Fei Tao,
and Qingsong Hua. “An Ontology-based Resource Reconﬁg-
uration Method for Manufacturing Cyber-Physical Systems.”
IEEE/ASME Transactions on Mechatronics (2018):7-14.
 Wan, Jiafu, Baotong Chen, Shiyong Wang, Min Xia, Di Li,
and Chengliang Liu. “Fog Computing for Energy-aware Load
Balancing and Scheduling in Smart Factory.” IEEE Transactions
on Industrial Informatics (2018): 223-229.
 Pillai, Padmanabhan, and Kang G. Shin. “Real-time dynamic
voltage scaling for low-power embedded operating systems.” In
ACM SIG-OPS Operating Systems Review, vol. 35, no. 5, pp.
89-102. ACM, 2001.
 Abolfazli, Saeid, Zohreh Sanaei, Ejaz Ahmed, Abdullah Gani,
and Rajkumar Buyya. “Cloud-based augmentation for mobile
devices: motivation, taxonomies, and open challenges.” IEEE
Communications Surveys and Tutorials 16, no. 1 (2014): 337-368.
 Fu, Xiong, Juzhou Chen, Song Deng, Junchang Wang, and
Lin Zhang. “Layered virtual machine migration algorithm for
network resource balancing in cloud computing.” Frontiers of
Computer Science 12, no. 1 (2018): 75-85.