Gravitational Emulation-Grey Wolf Optimization Technique For Load Balancing in Cloud Computing

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Gravitational Emulation-Grey Wolf Optimization

technique for Load balancing in Cloud Computing*

Kethavath Prem Kumar


Computer Science & Engineering
ACE Engineering college
Ghatkeshar, Hyderabad, India
kpremkumar2018@yahoo.com

Abstract—Cloud computing is a virtual pool of computing is NP-hard optimization problems like travelling salesman
resources such as software, platform, infrastructures, problem, combinatorial problems like integer programming
applications, storage and information provides to users through and addressing problems. These problems occur due to the
the internet. The computing capacity of the virtualized cloud allocation of hundreds and thousands of VMs to cloud
frame is divided into several VMs with the help of virtualization resources that causes delay in the performance of TS [5].
methodology. In the cloud environment, load balancing is the
challenging task because distribute the task equally among the In TS the datacenters, brokers, VM and cloudlets have been
VMs is a complex one. In this paper, hybrid Grey-Wolf- created for performing the required task using the CloudSim
Optimization (GWO) algorithm and Gravitational-Emulation- toolkit to satisfy customer’s demand [6]. CloudSim is a
Local-Search (GELS) algorithm is utilized for cloud load framework for simulation of cloud infrastructure, where
balancing. A GWO based optimization method is used for datacenters are resource provider, brokers help in creation and
ranking the host resources based on their efficiency and destruction of VMs, cloudlets perform as task submission and
utilization. The GELS algorithm is used for clustering the tasks provide the information about the RAM size, bandwidth and
based on their VM ability and cost. The proposed GWO-GELS number of processor allocations [7]. In cloud environment,
method significantly decrease the response time and cost. An resource distribution to equally control the loads is a very
experimental analysis is demonstrated that the makespan essential task. The resource allotment task performs after the
performance of proposed method achieved 10msec lesser than
process of TS. The incoming task requests are equally allocated
existing methods. The evaluated result validated that GWO-
to resources with the help of Load Balancing technique [8].
GELS performed effectively by means of energy consumption
minimization, reduction of cost, makespan and avoid Degree of
The traditional research work of cloud load balancing proposed
Imbalance (DI). The GWO-GELS method is implemented in several solutions. Each solution consists of specific methods
CloudSim. such as load balancing algorithms, round robin, Max-Min and
Min-Min algorithm, etc. But, an existing algorithm has some
Keywords—component; formatting; style; styling; insert (key limitations which becomes inefficient to handle the workload
words) in cloud computing [9]. The major responsibilities of load
balancing are to balance the work load of the whole system
I. INTRODUCTION while at the same time reduces the makespan of a set of tasks
[10, 11]. This research work, focused on two objectives such as
Cloud computing is fundamentally an internet dependent GWO algorithm is utilized with the intention of ranking up
service offers a cloud storage accessible only under through every host resources according to their efficiency and
subscription or pay per use scheme. A typical cloud utilization. As a result, it’s significantly reduced the makespan.
environment contains several fundamental components like Next, clustering every task into clusters suitable according to
Datacenter, Virtual Machine (VM), Cloudlet, Broker etc., different VM ability and cost, GELS are utilized.
among them Data center is the one of the most interesting
research areas that include a number of various domains like This paper is composed as follows. Section II presents a
computing, networking, management, and so on [1]. Recent broad survey of many recent papers of different load balancing
technologies and methodologies require to develop the speed of techniques in a cloud environment. In section III, the
the data capacity growth, and to deal with the underlying combination of GWO and GELS algorithm is described. In
system complexity, power consumption and etc. [2]. A data Section IV, comparative experimental result of existing and
center can be defined as a facility with all the resources proposed strategy is presented. The conclusion is made in
required for storage and processing of digital information [3]. Section V.
Data centers are critical to design in system architecture
because of some elements like processor, memory, and I/O II. LITERATURE REVIEW
devices are dynamically shared [4]. In cloud computing Researchers suggested several techniques in cloud
platform one of the major challenges faced by Cloud Service environment based load balancing. In this scenario, brief
Provider (CSP) is Task Scheduling (TS). Generally, issue in TS

978-1-5386-5657-0/18/$31.00 2018
c IEEE 177
evaluations of some important contributions to the existing theorem was applied in the clustering process to produce the
literatures presented below. best clustering set of physical hosts. An experimental analysis
demonstrated that the proposed method decreases the failure
S. L. Chen et al., [12] presented a dynamic annexed balance number of task deployment activities compared to the
technique to solve the issue of static load balancing and traditional methods. An advanced LB-BC method achieved
estimate the capacity of load balancing. The Cloud Load better results in external services of the cloud data center and
Balancing (CLB) was majorly concentrated on both computer improves the throughput. The method was only applicable to
loading as well as sever processing. By using CLB, a server Local Area Network (LAN).
would unable to handle excessive computational requirements.
An algorithm was applied to both virtual web servers and
physical servers. The results showed that cloud server III. LOAD BALANCING OF VMS USING PROPOSED GWO AND
performance based on the architecture balanced the loading GELS ALGORITHM
performance when users logged in at the same time. The The cloud load balancing is the significant task to be
algorithm was simple and very efficient compared to the considered in the cloud computing technique. The major
existing methods. The method suffered from the time objective of the load balancing is to distribute the local
complexity which was a major drawback. workloads equally, as a result no VM is overloaded or idle. The
L. Liu, et al., [13] proposed resource allocation based load balancing methods decrease the response time, makespan
multi-Quality of Services (QoS) Load Balance Resource and maximize the resource utilization in cloud environment. In
Allocation Method (MQLB-RAM) method. The resource this paper, majorly, two points are highlighted, initially, for
allocation was classified into two sections such as initially all reducing makespan purpose, GWO algorithm is utilized with
assigned virtual peers to physical hosts. The second section the intention of ranking host resources according to their
was the bound tasks assigned by particular sensor users to efficiency and utilization. The GELS algorithm is used for
virtual peers and the relative factors were decomposed in both clustering the task based on their VM ability and cost. The
parts. An experiments based on the CloudSim showed that the proposed architecture is shown in the figure.1. The five major
method MQLB-RAM implemented a reasonable resource steps of the proposed work are described in flowing sections.
allocation and achieved a good balance between load, resource Step 1: The GELS method used for initial population
access performance improvement and cost. The major generation and it consider the primary velocity to select the
disadvantage was the response time in the allocation method total population.
caused delays in case of a larger number of data center peers.
Step 2: The evaluation of fitness values of each wolf is
M. Vanitha, and P. Marikkannu, [14] presented a Dynamic generated by the GWO algorithm and these algorithms
Well-Organized Load Balancing (DWOLB) algorithm for generate the candidate response based on the load estimation of
balancing workloads and sufficient resource utilization. The VM.
load balancing employed Genetic Algorithm to distribute the
workload into virtual machine that were available in the Step 3: Calculate the mass and acceleration of the load by
network. The method balanced the loads on the various servers using these algorithms and update the position of the wolves
in order to avoid congestion in the network and to reduce the based on the acceleration.
consumption of the resources. The method allowed a greater Step 4: Global position updation is done by using
number of servers to be attached to the VM. The use of GA in acceleration. The formula is followed by the equation (1).
combination with DWOLB allowed the reduction of costs and
power consumption greatly. The fitness level declined with an Acceleration[i][ j] Force[i][ j] / mass[i] (1)
increase in the number of servers and this increased the CPU
utilization, since fitness depended on load and load balancing Step 5: If a number of iterations are executed then the
factors directly. algorithm generates the best solution. Assume that, the
iterations consist of minimum value represents the iteration
Gaochao Xu, et al., [15] presented a load balancing strategy keep on running till it reaches a maximum value.
was based on a public cloud. This method was employed in
cloud partitioning with the switch mechanism to choose the
different conditions. The game theory method was used in load
balancing technique to improve the efficiency in a public cloud
environment. The load balancing strategy was made cloud
computing more effective, efficient and user satisfaction level
was enhanced. When the cloud environment was huge and
composite then the proposed load balancing model was
partition the public cloud into several groups. So, these
partitions reduced the load balancing complexities. The method
was unable to test the system availability and efficiency which
was a major drawback of the scheme.
J. Zhao, et al., [16] presented Load Balancing based on
Bayes and Clustering (LB-BC) method based on task
deployment and employed for long term process. The Bayes

178 Second International Conference on Green Computing and Internet of Things (ICGCIoT)
population is generated and the position of every wolf is
initialized. The co-efficient vectors of A and C are described
in equation (2) and (3).
A 2a.r1  a (2)

C 2r 2 (3)

The A takes random values in the rage of > a, a @ ,


C with a random value in the range [0, 2] and it avoids the trap
of local optimal. In the traditional GWO algorithm, the vectors
linearly decrease from 2 to 0 during the execution of every
iteration. After the initialization of the coefficients, every wolf
(search agent) fitness value is estimated. After that, best fitness
solutions are selected as first, second and third such as α, β, and
δ, respectively.

( DD ) C1. X D  X ,(( DE ) C2 . X E  X ,(( DD ) C3 . X G  X ,


(4)

X1 X D  A1.(( DD ),
) X2 X E  A2 .(( DE )), X 3 X G  A3 .(( DG ),
)
(5)

x1  x2  x3
x(t  1)) (6)
3
In each iteration of the algorithm, the wolf’s position
update depends on the position of wolves α, β, and δ according g
to Equations (3), (4) and (5). In addition, values of vectors A ,
Fig. 1. Proposed method based cloud load balancing
C and a are updated. On the basis of new positions, the value
of the fitness function of wolves is calculated and D , E and G
will be selected. The GWO algorithm is used for ranking up
every host resource based on their efficiency and utilization. As
A. Grey Wolf Optimization Algorithm a result, the makespan decrease gradually.
In GWO, the optimization is done by meta-heuristic 1) Pseudocode for GWO algorithm
techniques, which is bio-inspired from nature of grey wolves. Initialization of grey wolf population X i (i=1, 2 … n)
In a grey wolf community there are four categories of grey
wolves namely alpha, beta, delta, and omega. Among them Initialization of A , C and a parameters
alpha is considered to be the leader of the group. Beta wolves
assist alpha in decision making and hunting which are Measure every agent or wolf fitness value
considered to be the next candidate eligible to be alpha if alpha
X G third best search agent
attains the stage of retirement or death while hunting. Delta
wolves are elder wolves or former alpha wolves or sentinels or X E Second best search agent
scout that protects the boundaries of their group. Omega
wolves are the least prioritized wolves which need to submit to X G third best search agent
all the other wolves and follow all other category wolves.
Assume that every wolf is searching solution in the while(t  max number of iterations)
search space. The wi ^wi1 , wi 2,....... win ` represents position For each search agent
vectors in the search space, whereas the dimension of the Update the position of current search agent
problem is shown as n The fitness function (based on End for
problem definition) is employed to estimate the position of the
wolves. Based on the fitness value the best wolves are update a, A, and C
classified as first solution that is represented as D , second is E ,
Calculate fitness of all search agents
and third is G respectively. In the best solution searching
process, the wolves update their position according to the update XD , X E , X G
position of D , E and G . In the starting stage, the wolf

Second International Conference on Green Computing and Internet of Things (ICGCIoT) 179
t t 1 ™ Iterations: The algorithm iterations are defined and
before terminate the algorithm all iterations are
end while executed.
return X D ™ Pointer: Pointer is employed to detect the direction of
movement of the elements in the vectors.

Furthermore, the swarm intelligent methods are usually C. Pseudocode for proposed GWO and GELS Algorithm
employed to solve the optimization problems that can’t have begin
the leader to monitor the entire proceeding period. This
limitation is resolved in GWO method; the grey wolves have Initialize VMs randomly
individual leadership capacity. Moreover, this algorithm uses
Evaluate the fitness for each wolf
few parameters and implementation is simple.
Assign a predefined starting solution as current solution
B. Gravitational Emulation Local Search Algorithm
Assign an initial velocity randomly in the dimension
In 2004, Barry Webstar proposed GELS method, which
depends on gravitational attraction and its repeats the process Calculate an initial vector velocity sum
of nature for searching the best solution in a search space. The Best _ solution current _ solution
two objects are near to each other and strong gravitational force
works between them. For the reason of gravity two objects pull while(velocity _ sum! 0 OR i  max_ITER)
towards each other. The algorithm describes the randomization
along with two essential parameters such as gravity and Do
velocity. The equation (7) describes the gravitational force of
Newton’s formula. velocity _ sum 0

Gm1.m2 Generate candidate response


F (7)
R2 Estimate Mass and Acceleration

Where, (6.672 universal constant) is gravitational Analyze the difference in GF between current
constant, the first and second object masses are m1& m2 . Estimate the Solution and candidate solution with the help
R represents the distance between two objects. The GELS of equation (8)
method includes two steps for pointer movement in the search
space. In the beginning stage, a candidate solutions are selected if (fit(candidate solution) > fit(current solution) )
based on the current response of neighborhood. In second, then
movements are allowed outside the neighborhood of current
responses. In GELS, formula in equation (7) is modified and Best _ solution candidate _ solution
the gravitational force has been calculated as follows:
Update _ Velocity(v, force)
G CU  CA
F (8) end
R2
Global position update using acceleration
Where, CU is the current response, CA is Candidate
response, the variable R is constant or may change in each Return best solution found
iteration, and G (6.672) represents Gravitational constant. The end
mass is replaced by difference of Current response and
Candidate response. The above equation is designed to be a The VMs are initialized randomly and GELS algorithm is
positive value indicates an objective function value of current used to select the total population and high primary velocity is
solution is larger than that of the candidate solution. The taken into concern. The equation (8) shows the gravitational
negative value represents the only candidate is larger. force of present response and candidate response. After that,
it’s applied to the velocity of that dimension of the candidate
1) Parameters used in Gravitational Emulation Local response. In evaluation step, calculate the fitness of every wolf
Search Algorithm using GWO method. Here, calculates the alpha, beta, delta and
™ Max velocity: Defines the maximum value that any omega parameters. After that, with the help of GELSA
element within the velocity vector can have used to calculates the mass and acceleration. Moreover, all global
prevent velocities that became too large to use. positional are updated using acceleration and finally, obtain the
™ Radius: The radius values are adjusted for best solution.
gravitational force formula and decide how fast the
gravitational force increase or decrease. IV. EXPERIMENTAL RESULT AND DISCUSSION
For experimental simulation, CloudSim 3.0.3 PlanetLab
workload was employed on PC with 3.2 GHz with i5

180 Second International Conference on Green Computing and Internet of Things (ICGCIoT)
processor. The GWO-GELS algorithm performance was server. The table.1 represents the performance of proposed load
compared with the traditional load balancing techniques such balancing technique.
as Ant Colony Optimization (ACO), Particle Swarm
Optimization (PSO), Honey-Bee-Behavior Load Balancing TABLE I. PERFORMANCE OF THE PROPOSED LOAD BALANCING
(HBB-LB), and Energy-aware Fruitfly optimization for load TECHNIQUE
balancing (EFOA), and Simulated annealing based FOA for
load balancing (FOA-SA-LB). Number of Makespan Response DI Cost($)
Task Time
10 2.07 100.2 1.278 0.186
A. Evaluation Metrics 20 4.07 200.2 1.633 0.373
The proposed GWO and GELS method performance is 30 5.27 240.31 1.825 0.6
40 7.13 33.53 1.6855 0.786
measured using different evaluation metrics such as execution
50 8.00 413.53 1.727 0.973
time, overall cost, and DI are described below. 60 10.2 466.98 1.874 1.2
70 12.07 530.2 1.881 1.386
1) Degree of Imbalance
80 14.07 610.2 1.801 1.573
The VM overloaded tasks, resource usage of RAM, MIPS 90 15.13 730.2 1.8919 1.8
and bandwidth and cloudlet traffic creates the DI in cloud 100 17.13 810.2 1.8950 1.98
environment. The DI is used to measure imbalance among
VMs to spread workload evenly for VM and its described in
equation (9), The value of GWO-GELS load balancing technique’s
performance parameters makespan, response time and DI
T max T min preference is shown in table 1. An existing ACO based GELS
DI (9)
Tavg method achieves less results, compared to GWO-GELS
method [17]. The makespan performance is shown in table.2
Where Tmax and Tmin are highest and shortest time span and figure.2. When the number of tasks increases, the DI in the
took for a cloudlet to process in proposed framework and Tavg system becomes limited, energy consumption minimizes,
response time becomes faster, and cost also minimize.
is average time span took by system to process one cloudlet.
Thus DI will interpret balance of workload among different
VMs in order to equally distribute working time for overall TABLE II. EXISTING AND PROPOSED MAKESPAN PERFORMANCE
task. Number Proposed PSO [18] HBB-LB EFOA FOA-
of Task GWO- [19] SA-LB
2) Execution time TE GELS
If workload of j th task which is hosted on i th resource of 10 2.07 8.987 8.254 7.147 5.213

Wj 20 4.07 10.846 10.268 9.954 8.543


processing capacity Pi then execution time of j th task is . 30 5.27 18.923 17.234 12.567 11.213
Pi
Hence, 40 7.13 20.896 19.124 15.368 12.547
50 8.00 28.236 20.458 19.254 15.123
§ Wj · 60 10.2 30.786 25.147 20.741 18.234
TE ¨ ¸ * X ji (10)
© Pi ¹ 70 12.07 34.256 28.364 22.876 21.147
80 14.07 42.863 31.587 27.124 21.785
It is assumed that the execution time is proportional to
execution cost. The X ji indicates the binary variable for the 90 15.13 49.256 37.254 30.247 23.895

position of the task j and if the task was hosted on the 100 17.13 55.781 45.189 31.014 27.124

resource i .
3) Energy Utilization The table 2 represents the makespan performance of
In cloud architecture, to avoid the wastage of static power proposed and existing methods. Here, consider the number of
dissipation from host machines, the servers which are in idle tasks 10 to 100 ad number of tasks, the difference in makespan
state are changed to sleep condition. By doing this process, the time is quite high. Compare to the existing PSO, HBB-LB,
energy consumption of the cloud is lessened. In the proposed EFOA and FOA-SA-LB technique, proposed GWO-GELS
framework, the power consumed by a server is described with method takes less task completion time.
the help of equation (11) as, The figure 2 shows the makespan performance of existing
P(u) ku Pmax  (1  k) u Pmax u u (11) optimization methods such as PSO, HBB-LB, EFOA-LB and
proposed GWO-GELS method. The x-axis represents the
Where, P(u) is denoted as single server power utilization, number of tasks and y-axis represents the makespan in msec.
The result demonstrated that proposed GWO-GELS has
Pmax is represented as maximum power utilized by a server and minimum makespan time compared to the traditional methods.
k is denoted as the proportion of power utilized by an idle
TABLE III. EXISTING AND PROPOSED DI PERFORMANCE

Second International Conference on Green Computing and Internet of Things (ICGCIoT) 181
Number of Proposed ACO [18] LBACO [20] related optimization technique minimize the DI. Compared to
Task Method GWO- the existing methods like ACO and Load Balancing ACO
GELS
(LBACO) the DI in load between VMs are maximized. The
100 1.22 2.389 2.356
200 1.25 2.284 2.195
proposed method checks the VMs capacity after that assign the
300 1.23 2.541 2.132 workloads to each VM. Moreover, the proposed method
400 1.24 2.787 2.01 significantly balances the workloads in cloud environment.
500 1.25 2.689 1.986

The table 3 represents the DI of proposed and existing


optimization technique’s performance. The load balancing

Fig. 2. Makespan

Fig. 3. Degree of Imbalance

The DI of proposed and existing algorithm with the number Number of Proposed FOA- EFOA- HBB-LB PSO
of tasks are varying from 100 to 500 is shown in Figure.3. It is tasks GWO- SA- LB
GELS LB
clearly evident that the DI performance of the ACO and
100 0.82 1.13 1.17 2.44 1.78
LBACO algorithm is maximum than GWO-GELS algorithm.
If the amount of tasks increase, then the VMs workloads 200 0.87 1.20 1.28 3.66 3.12
become imbalanced. The proposed GWO-GELS method 300 1.09 2.01 2.21 4.44 4.34
maintains the best results in different number of tasks. 400 1.23 3.08 3.20 5.5 7.21
Table.4 Energy consumption of proposed and existing load
balancing techniques The table 4 represents the energy utilization of proposed
and existing load balancing methods in cloud environment. The

182 Second International Conference on Green Computing and Internet of Things (ICGCIoT)
GWO-GELS method reduce the power utilization in data
centers to decrease the cost. Compare to the traditional
optimization based load balancing technique the proposed
method uses less power and decrease the cost.

Fig. 4. Energy utilization

The figure 4 depicts the energy consumption performance consolidation of virtual machines,” Computing, vol. 98, no. 6, pp. 641-
of the proposed and existing load balancing methods. The 660, 2016.
proposed method balances the workloads and reduces the [3] S. Vakilinia, “Energy efficient temporal load aware resource allocation
in cloud computing datacenters,” Journal of Cloud Computing, vol. 7,
energy consumption significantly. When no tasks are assigned no. 1, p.2, 2018.
to the VMs then the proposed method helps to turn off the VMs [4] V. Murudi, K.M. Kumar, and D.S. Kumar, “Multi Data Center Cloud
(sleep state). This strategy reduces the energy consumption and Cluster Federation-Major Challenges & Emerging Solutions”, in IEEE
cost of the datacenter. Compared to the existing methods the International Conference on IEEE,Cloud Computing in Emerging
proposed method uses minimum energy consumption and Markets (CCEM), pp. 107-112, 2016.
better results. [5] L. Wang, Y. Yin, and Y. Zhong, “Cuckoo search with varied scaling
factor,” Frontiers of Computer Science, vol. 9, no. 4, pp. 623-635, 2015.
[6] X. Geng, Y. Mao, M. Xiong, and Y. Liu, “An improved task scheduling
V. CONCLUSION algorithm for scientific workflow in cloud computing environment,”
In this paper, a hybrid GWO-GELS method is utilized for Cluster Computing, pp.1-10. 2018.
cloud load balancing and it’s implemented for solving the load [7] S.H.H. Madni, M.S.A. Latiff, and Y. Coulibaly, “Resource scheduling
balancing problems in cloud. The proposed GWO algorithm for infrastructure as a service (IaaS) in cloud computing: Challenges and
opportunities,” Journal of Network and Computer Applications, vol. 68,
provides the ranks to each host resource based on their pp. 173-200, 2016.
efficiency and utilization and GELS algorithm clusters the task [8] S.P. Praveen, K.T. Rao, and B. Janakiramaiah, “Effective Allocation of
based on VM ability and cost. The combination of these two Resources and Task Scheduling in Cloud Environment using Social
algorithms essentially reduces the response time and cost. An Group Optimization,” Arabian Journal for Science and Engineering,
experimental analysis demonstrated that the proposed GWO- pp.1-8, 2017.
GELS method achieved minimum DI, makespan, energy [9] F. Vhansure, A. Deshmukh, and S. Sumathy, “Load Balancing in Multi
utilization, and cost in cloud data centers. The proposed Cloud Computing Environment with Genetic Algorithm,” In IOP
Conference Series: Materials Science and Engineering, IOP Publishing,
method achieved better results than existing methods. In future, vol. 263, no. 4, p. 042010, 2017.
the work can be extended as Quality of Service in load [10] M. Aruna, D. Bhanu, and S. Karthik, “An improved load balanced
balancing like network traffic information in cloud computing. metaheuristic scheduling in cloud,” Cluster Computing, pp.1-9, 2017
In the cloud, the network traffic is one of the major problems, [11] N.J. Kansal, and I. Chana, “An empirical evaluation of energy-aware
which consumes maximum power and increase the datacenter load balancing technique for cloud data center,” Cluster Computing,
cost. To avoid this problems research work further focuses on pp.1-19, 2017.
load balancing with QoS in a cloud environment. [12] S.L. Chen, Y.Y. Chen, and S.H. Kuo, “CLB: A novel load balancing
architecture and algorithm for cloud services,” Computers & Electrical
Engineering, vol. 58, pp. 154-160, 2017.
REFERENCES [13] L. Liu, H. Mei, and B. Xie, “Towards a multi-QoS human-centric cloud
[1] S. Ismaeel, A. Miri, and A. Al-Khazraji, “Energy-consumption computing load balance resource allocation method,” The Journal of
clustering in cloud data centre”, in 3rd MEC International Conference on Supercomputing, vol. 72, no. 7, pp. 2488-2501, 2016.
IEEE, Big Data and Smart City (ICBDSC), pp. 1-6, 2016. [14] M. Vanitha, and P. Marikkannu, “Effective resource utilization in cloud
[2] L. Salimian, F.S. Esfahani, and M.H. Nadimi-Shahraki, “An adaptive environment through a dynamic well-organized load balancing
fuzzy threshold-based approach for energy and performance efficient

Second International Conference on Green Computing and Internet of Things (ICGCIoT) 183
algorithm for virtual machines,” Computers & Electrical Engineering, [18] M. Lawanyashri, B. Balusamy, and S. Subha, “Energy-aware hybrid
vol. 57, pp. 199-208, 2017. fruitfly optimization for load balancing in cloud environments for EHR
[15] G. Xu, J. Pang, and X. Fu, “A load balancing model based on cloud applications,” Informatics in Medicine Unlocked, vol. 8, pp. 42-50,
partitioning for the public cloud,” Tsinghua Science and Technology, 2017.
vol. 18, no. 1, pp. 34-39, 2013. [19] P.V. Krishna, “Honey bee behavior inspired load balancing of tasks in
[16] J. Zhao, K. Yang, X. Wei, Y. Ding, L. Hu, and G. Xu, “A heuristic cloud computing environments,” Applied Soft Computing, vol. 13, no.
clustering-based task deployment approach for load balancing using 5, pp. 2292-2303, 2013.
Bayes theorem in cloud environment,” IEEE Transactions on Parallel [20] [20] Li, K., Xu, G., Zhao, G., Dong, Y. and Wang, D., 2011, August.
and Distributed Systems, vol. 27, no. 2, pp. 305-316, 2016. Cloud task scheduling based on load balancing ant colony optimization.
[17] J. Yadav, and S. Tyagi, “Hybrid of Ant Colony Optimization and In Chinagrid Conference (ChinaGrid), 2011 Sixth Annual (pp. 3-9).
Gravitational Emulation Based Load Balancing Strategy in Cloud IEEE.
Computing,” International Research Journal of Engineering and
Technology (IRJET), vol. 04, no. 07, 2017.

184 Second International Conference on Green Computing and Internet of Things (ICGCIoT)

You might also like