Materials Today: Proceedings xxx (xxxx) xxx

Contents lists available at ScienceDirect

Materials Today: Proceedings

journal homepage: www.elsevier.com/locate/matpr

An energy efficient load balancing on cloud computing using adaptive

cat swarm optimization
K. Balaji a,⇑, P. Sai Kiran b, M. Sunil Kumar c
a
Department of CSE, Koneru Lakshmaiah Education Foundation, Vaddeswaram, AP, India
b
Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, AP, India
c
Department of Computer Science and Engineering, Sree Vidyanikethan Engineering College(Autonomous), Tirupati, AP, India

a r t i c l e i n f o a b s t r a c t

Article history: Cloud has been a dominant player in information technology in recent years. To meet the growing
Received 30 October 2020 demands for computing and storage resources, cloud providers deploy data centers (DC) across the globe
Accepted 6 November 2020 that includes hundreds of thousands of IT servers. Due to massive energy use, every DC leave a huge car-
Available online xxxx
bon footprint and thus, have bad effects on the environment. Because of this, efficient computing
resource utilization and minimization of energy usage are critical issues to create cloud computing more
Keywords: booming. Efficient load balancing is the primary means of solving these problems. In order to overcome
ACSO
optimization problems, this study has developed a load balancing system using adaptive cat swarm opti-
Load balancing
Energy consumption
mization (ACSO) algorithm. The efficiency of the proposed technique is analyzed by using different value
Cloud computing metrics and performance differentiated with other methods.
Virtual machine Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the Emerging Trends in
Materials Science, Technology and Engineering.

1. Introduction Adaptation and development of cloud computing security

issues are the most important barrier [11,12]. Cloud Client applica-
Cloud computing resources are to provide high-speed internet tion to access the services of any key concern for the safety and risk
services to users through an application [1]. Cloud provides hosting [13]. Cloud computing is a wide area. It is widely applied in the
and storage services on the Internet, which is the practical evolu- current period, so that load balancing has become a major issue
tion of technologies [2]. This requirement, virtualized, location [14]. Load balancing is one of the key issues of the cloud, which
independent, paid the price model, which is aimed at achieving evenly maintains jobs across different VM’s to provide the right
optimal resource utilization and high performance [3,4]. But the IT resources to improve comprehensive system performance
QoS, load balancing resource utilization, fault tolerance, security [15,16]. It offers low energy consumption and carbon emissions
these are some complications. It has gained great popularity in with the low rates, which supports to accomplish green computing
recent years [5–7]. The use of cloud computing services for a per- [17]. Users according to their needs and resources, the demand for
sonal or corporate level. Cloud services can be summed up as a the service, which provides access to information and software.
model, a resource that is accessible with less management effort With the increased popularity, the biggest broaden in the require-
[8]. Clouds can be categorized into private, public and hybrid based ments of the users of the services, this can be accomplished with
on their architecture [9]. It provides three types of services such as efficient load balancing methods [18]. Load balancing confess the
a Infrastructure as a Service (IAAS), which provides a demanding sharing of resources among different nodes in the cloud and is
infrastructure. Software as a Service (SAAS) provides software ser- aimed at the efficient use of services, improving system perfor-
vices. Platform as a Service (PAAS), it contributes platforms for pro- mance, and reducing energy utilization. As a result of reduced
gram advancement [10]. energy usage. Many encryption mechanisms store data in the cloud
to protect the load balance [19,20].
In this article, we proposed an algorithm to increase the system
resource usages of the VM and reduce power usage. Then, the per-
⇑ Corresponding author. formance of the proposed model is checked with the help of differ-
E-mail address: [email protected] (K. Balaji). ent metrics. The beneficence of the work is presented here;

https://doi.org/10.1016/j.matpr.2020.11.106
2214-7853/Ó 2021 Elsevier Ltd. All rights reserved.
Selection and peer-review under responsibility of the scientific committee of the Emerging Trends in Materials Science, Technology and Engineering.

Please cite this article as: K. Balaji, P. Sai Kiran and M. Sunil Kumar, An energy efficient load balancing on cloud computing using adaptive cat swarm opti-
mization, Materials Today: Proceedings, https://doi.org/10.1016/j.matpr.2020.11.106
K. Balaji, P. Sai Kiran and M. Sunil Kumar Materials Today: Proceedings xxx (xxxx) xxx

For load balancing ACSO algorithm is developed for reducing In equation (1), the first term is represented as power consump-
the energy utilization. tion (P), the second term is represented as migration cost (MC) and

The performance is analyzed using make span, and energy uti- third term is represented as memory utilization (MU).
lization and memory utilization. Power consumption is the principal parameter of load balanc-
ing process. The Power consumption of the framework is deter-
The remaining paper is explained in the following manner, the mined by the absolute Euclidean distance (ED) of all the dynamic
literature based research papers are analyzed in part 2 and fitness PM simultaneously. The smaller ED is considered as the better load
function design is given in part 3. The proposed methodology is balanced system. The PM is turned off, when no assignment is run
discussed in part 4 and the experimental results analyzed in part in relating PM. The Power efficient factor (EF) of each active node is
5. The conclusion part is presented in 6. calculated based on equation (2).
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u d
uX
EF ¼ t
2
2. Literature review ðU i  Ubest i Þ ð2Þ
i¼1

A lot of researchers developed a load balancing methods. Out of

Where,
this, very few works are studied here; Li, C., et al. [21] have inves-
i ? Resources namely CPU and memory
tigated a heap adjusting based cross-layer in the cloud. The pro-
U i ? Present resource utilization
posed method expands neighborhood cloud service pools with
Ubesti ? Best utilization of resource i for power efficiency in
the public cloud to build the likelihood of meeting the service level
each physical node.
agreements (SLA). The test result shows the heap adjusting based
power efficiency at time t is calculated as follows;
cross-layer was superior to the next enhancement procedure.
X
Moreover, Al-Mashhadi, S., et al. [22] have proposed a protected Pt ¼ EF ð3Þ
load balance in the cloud. These methodologies incorporate utiliz-
ing new system topology, server and system virtualization, utiliza- The total power efficiency of the system within T is given by (4)
tion of programming characterized systems innovation and system
X
T
load adjusting. Be that as it may, since SDN normally deals with all P¼ Et ð4Þ
system assets from a focal area, more looks into are required to t¼0
make sure about SDN correspondence between its various layers.
Migration cost is the another factor for the objective function.
Galloway, J.M., et al. [23] have analyzed force aware load adjust-
The MC of VM is expanded when the quantity of movements incre-
ing in the cloud. Right now, present a load adjusting way to deal
ments. The great load balancing framework ought to keep up least
with IaaS cloud designs that were power mindful. Since the cloud
movement. The computation of the MC of the whole Cloud
engineering executed by neighborhood associations will, in gen-
arrangement is determined utilizing condition (5).
eral, be heterogeneous, they consider in the structure. Force Aware
V  
Load Balancing calculation, PALB, keeps up the condition of all reg- 1X Number of migrations in VMs
ister hubs, and dependent on usage rates, and chooses the number MC ¼ ð5Þ
V i¼1 Total number of VMs
of figure hubs that ought to be working. They show that the
arrangement gives sufficient accessibility to figure hub assets Memory Usage is the another factor of the load balancing objec-
while diminishing the general force devoured by the nearby cloud tive function. Memory is only a heap. The heap structure legiti-
by 70%97% contrasted with utilizing load adjusting procedures mately relies upon the benefits utilized by the VM for setting up
that were not power-aware. the assignments from different customers. Various resources uti-
In, [24], Pandey, S., et al., has built up a problematic system for lized by the VM are CPUs and memory storages. The computation
load adjusting in the cloud. That paper improves way sharing at the of the storage usage of the whole Cloud arrangement is determined
time of movement during the load balance process. A cloud work- utilizing condition (6).
load organize was developed dependent on Greedy heuristics hav- " !#
ing the attributes to construct the steadily various set, choosing 1 XPM XVM
1 CPU utilizedij memory utilizedij
MU ¼ þ
each process hub in turn to amplify a decent variety measure. PM  VM i¼1 j¼1
2 CPUij memoryij
The virtual machine director can depend on the principles and
ð6Þ
can settle on a productive choice for ideal provisioning of virtual
machine dependent on the totaled course. In equation (1), the objective function of our research is speci-
Similarly, Agrawal, R., et al. [25] has dissected a load adjusting fied. In this paper, the above objective function is going to be min-
in cloud processing. Right now, the ant colony optimization algo- imized using ACSO algorithm.
rithm has been utilized. Right now was a goal to develop such a
load adjusting technique that can improve the execution of cloud 4. Proposed energy efficient load balancing methodology
figuring close by most outrageous resource utilization.
Load balancing is mainly used to provide high availability in the
3. Fitness function design cloud environment, hardware maintenance and workload balance
and fault acquisition. In the load balancing process, the VM is
The main goal of proposed technique is to obtain finer load bal- transferred from one physical server to another physical server.
ancing on cloud. Basically, cloud computing contains u number of The load balancing approach reduces time consumption, energy
PMs and each PM consist of v number of VMs. The purpose of loss and delay time. Here, the adaptive cat swarm optimization
the proposed algorithm is to reduce Power (P), migration cost algorithm is used for load balancing and for the security process,
(MC) and Memory Utilization (MU) while the load is balanced. the refinement algorithm is used. The workflow diagram of the
Based on the above condition, the objective function is designed implemented model is shown in Fig. 1.
which is given in equation (1). For example, consider, the solution consists of four PM and each
Fitness Function ¼ Min½a1ðP Þ þ a2ðMCÞ þ a3ðMUÞ ð1Þ PM consists of three VM. At first, the incoming tasks are randomly
assigned to VM. And also, all the VM should balanced. If it is any
2
K. Balaji, P. Sai Kiran and M. Sunil Kumar Materials Today: Proceedings xxx (xxxx) xxx

Fig. 1. Flow diagram of implemented method.

PM1 PM2 PM3 PM4

1 1 0 1 1 1 1 0 1 1 1 1

Fig. 2. Initialization.

overloaded PM available means it should be migrated. The initial R = a random number between 0 and 1.
solution format is given in Fig. 2.
Step 2: Fitness calculation: After initialization, fitness value is
By using fitness calculation (FC) value we compare all the cat
calculated for each initialization condition. In this paper, we design positions. if all are equal set the identifying probability value
a multi-objective fitness function which is a combination of as 1 for all cats. Otherwise we can use below identifying prob-
energy, cost and memory utilization. ability equation (9).

Fitness Function ¼ Min½k1ðPÞ þ k2ðMC Þ þ k3ðMU Þ ð7Þ 

jFCi FCj 
Pv ¼ o<i<j ð9Þ
Here, the objective function is based on the three parameters jFCmax  FCmin j
such as energy (E), migration cost (MC) and memory utilization
Here,
(MU).
pv = probability of latest cat;
Step 3: Updation using adaptive cat swarm optimization
FCi = fitness calculation value of each cat;
After the fitness calculation, the solutions are efficient with the
FC max = maximum assessment of fitness calculation
help of ACSO. The ACSO has main two processes namely, seeking
FC min = minimum significance of fitness calculation
mode, tracing mode.
Tracing mode (TM): After seeing any pray cats start moving
Seeking mode (SM): After the fitness calculation, cats are given
from seeking mode to tracing mode. At the point when cats are
in seeking mode. Here, the cats are looking around the position and
in tracing mode, their conduct will be recreated with ACSO algo-
they adjust their position a little bit. In this, the cats are seeking an
rithm, and each cat follows the best position to update its own
optimal opportunity to find out the prey. The ACSO algorithm uses
velocity of cat n in dimension d. Tracing mode based updation is
four parameters: seeking memory pool (MP), seeking collection of
given in equations 10 and 11.
selected dimension (SR), counts of different dimensions (CD) and
self position consideration (SP) [26]. The procedure of seeking Vn; d ¼ Vn; d þ k  rðP best; d  Pn; dÞ ð10Þ
mode is explained below;
Here,

Initially, each cat duplicates its own position seeking memory Vn,d = Velocity of cat n in dimension d;
pool (MP) times. k = random number [0,1]

Then, these duplicated positions calculate count of dimensions r = constant;
to change (DC) using below equation(8) P best,d = location of the cat with best result;
Pn,d = current place of the cat.
The latest place of each cat is calculated with this equation (11)
Scn ¼ ð1  SR  RÞ  Sn ð8Þ
L n,d, new = L n,d, old + Vn,d
Here, in which,
Sn = recent position; L n,d, new latest place of cat n in dimension and
scn = Latest position; L n,d, old recent place of cat.

3
K. Balaji, P. Sai Kiran and M. Sunil Kumar Materials Today: Proceedings xxx (xxxx) xxx

16500
40
16000
Memory (bits)

Iterations
15500 30
15000 ACSO GA
20
14500 CSO CSO
14000 GA 10 ACSO
13500
0 10 20 30 40 0 2 4 6 8 10 12
Iterations Energy consumption (%)

Fig. 3. Performance analysis based on memory utilized. Fig. 4. Performance analysis based on energy consumption.

Algorithm 1: ASCO algorithm for load balancing

600
ACSO

Makespan (sec)
Begin
400 CSO
Inputs: cat count Cj (j = 1, 2, . . ., n), r, and SP
while (the termination is not fit) 200
calculate the fitness function for all cats
Cb = best solution for cat 0
10 20 30 40 50 60 70 80 90 100
while j > M
Number of task
If SP = 1
begin SM Fig. 5. Performance analysis based on Makespan.
Else
begin TM
End If
End while Energy consumpon 40
End While 30 ACS
Identify better outcomes O
20
End
(%)

10
0
Termination criteria: This mechanism terminates only when the 10 20 30 40 50 60 70 80 90 100
highest number of iterations is accomplished and the improved best Number of Tasks
fitness value is chosen, and this good solution based VM are
arranged in PM. Fig. 6. Comparison of energy consumption of the system by varying number of
tasks.

5. Result and discussion

The proposed methodology experimental results are examined

Similarly, proposed method taken minimum time of 430sec
in this section and performance are discussed in terms of various
which is 470sec for using CSO based load balancing and 490sec
assessment metrics. The proposed approach is implemented in
for GA based load balancing. In Fig. 6, the number of VM is set to
Cloudsim tool.
50, and the number of tasks increases from 10 to 100 in the gap
of 10. Here, the number of task increases, energy consumption also
5.1. Experimental results
increases. From this analysis, it’s clearly understood; our algorithm
got better performance compared to other methods.
The aim of this study is to reduce the energy consumption in the
VM’s in PM. To avoid the energy consumption and Makespan, load
balancing process is used in VM. To provide evidence of proposed 6. Conclusion
algorithm compared with different method namely, CSO based
load balancing and GA based load balancing. Energy efficient secure virtual machine migration using optimal
The performance of proposed methodology is inspected by cryptography methodology has been introduced in this paper. For
using memory utilized is shown in Fig. 3. In Fig. 3, our defined security purpose, the data migrated data encrypted with the help
method make use of the minimum memory compared to existing of sanitization method. Here, the load balancing has been devel-
methods. The minimum memory utilization is considered as the oped using ACSO algorithm. The mathematical explanation of
best load balancer. ACSO and sanitization algorithms has been explained properly.
Moreover, in figure, performance of our technique is inspected The experimental results are illustrated in terms of different met-
by using energy consumption. In Fig. 4, our method attain the min- rics. Compared to the existing algorithm, our proposed method
imum energy consumption of 6.1% which is 8.35 for using CSO attain the minimum make span and energy consumption. In future,
based load balancing and 8.9% for GA based load balancing. This we will implement our methodology with real time and also we
is due to ACSO. In Fig. 4, the p performance of our technique is will concentrate more.
inspected by using Makespan. As the number of tasks increases;
the Makespan gradually increases. When the task is set to be 20, CRediT authorship contribution statement
our proposed method taken 178sec which is 198sec for using
CSO based load balancing and 201sec for GA based load balancing K. Balaji: Conceptualization, Data curation, Methodology, Soft-
(Fig. 5). ware. P. Sai Kiran: Investigation, Visualization, Writing - original
4
K. Balaji, P. Sai Kiran and M. Sunil Kumar Materials Today: Proceedings xxx (xxxx) xxx

draft. M. Sunil Kumar: Software, Supervision, Validation, Writing - [12] B. Radojevic, M. Zagar ‘‘Analysis of issues with loadbalancing algorithms in
hosted (cloud) environments” Inproceedings of 34th International Convention
review & editing.
on MIPRO,IEEE, pp. 23-27 May 2011.
[13] T. Gunasekhar, K.T. Rao, V.K. Reddy, P. Sai Kiran, B. Thirumala Rao, ‘‘ Mitigation
of insider attacks through multi-cloud”, Int. J. Electric. Comput. Eng., (pp.136-
Declaration of Competing Interest 141).
[14] K. Radha, B. Thirumala Rao, S.M. Babu, K. Thirupathi Rao, V. Krishna Reddy, P.
The authors declare that they have no known competing finan- Saikiran. 2014. Allocation of resources and scheduling in cloud computing
with cloud migration. Int. J. Appl. Eng. Res. Volume 9, Issue 19, 2014, Pages
cial interests or personal relationships that could have appeared 5827-5837.
to influence the work reported in this paper. [15] S.T. Maguluri, R. Srikant, L. Ying. (2012). Stochastic models of load balancing
and scheduling incloud computing clusters. In 2012 INFOCOM (pp. 702–710).
IEEE.
References [16] C. Thiam, G.D. Costa, J.M. Pierson. (2013). Cooperative scheduling anti-load
balancing algorithmfor cloud: CSAAC. In IEEE international conference on
[1] V. Kunwar, N. Agarwal, A. Rana, J.P. Pandey, Load balancing in cloud—A cloud computing technology and science (pp.433–438).
systematic review, Big Data Analyt. (2017) 583–593. [17] S. Wang, K. Yan, W. Liao, S. Wang. Towards a load balancing in a three-level
[2] Z. Goudarzi, A. Faraahi, Effective load balancing in cloud computing, Int. J. cloudcomputing network. In: Proceedings of the 3rd IEEE International
Intell. Inf. Syst. 3 (6) (2017) 1. Conference on Computer Science and Information Technology (ICCSIT),
[3] S. Chhabra, A.K. Singh, Dynamic hierarchical load balancing model for cloud Chengdu, China, pp. 108–113 (2010).
data center networks, Electron. Lett. 55 (2) (2019) 94–96. [18] M. Randles, D. Lamb, A. Taleb-Bendiab, A comparative study into distributed
[4] K. Balaji, P. Sai kiran, 2017. Efficient resource allocation algorithm with optimal loadbalancing algorithms for cloud computing, in: Proceedings of 24th IEEE
throughput in cloud computing. J. Adv. Res. Dyn. Control Syst. Volume 9, 2017, International Conference on Advanced Information Networking and
Pages 1902-1910. Applications Workshops, 2010, pp. 551–556.
[5] L. Hong, G. Yufei. GACA-VMP: Virtual Machine Placement Scheduling in Cloud [19] K. Nishant, P. Sharma, V. Krishna, C. Gupta, K.P. Singh, N. Nitin, R. Rastogi.
Computing Based on Genetic Ant ColonyAlgorithm Approach, IEEE 12th Intl Loadbalancing of nodes in cloud using ant colony optimization. In:
Conf on Ubiquitous Intelligence and Computing and 2015 IEEE 12th Intl Conf Proceedings 14th InternationalConference on Computer Modelling and
on Autonomicand Trusted Computing and 2015 IEEE 15th Intl Conf on Scalable Simulation (UKSim), pp. 3–8. IEEE (2012).
Computing and Communications and Its Associated Workshops(UIC-ATC- [20] S.G. Domanal, G. Ram Mohana Reddy. Load balancing in cloud computing
ScalCom), 16158285 ,2015. using modified throttled algorithm. In: IEEE, International conference on CCEM
[6] K.S. Swami, P. Sai Kiran. 2018. Secure data duplication with dynamic (2013).
ownership management in cloud storage. J. Adv. Res. Dyn. Control Syst., [21] C. Li, L. Li, Load-balancing based cross-layer elastic resource allocation in
Volume 10, Issue 12 Special Issue, 2018, Pages 753-761. mobile cloud, Wireless Pers. Commun. 97 (2) (2017) 2399–2437.
[7] N. Chandrakala, B. Thirumala Rao, Migration of virtual machine to improve the [22] S. Al-Mashhadi, M. Anbar, R.A. Jalal, A. Al-Ani. 2020. Design of Cloud
security in cloud computing, Int. J. Electric. Comput. Eng. 8 (1) (2018) 210–219. Computing Load Balance System Based on SDN Technology. In
[8] K. Ramana, A. Subramanyam, A. Ananda Rao, Comparative analysis of Computational Science and Technology (pp. 123-133). Springer, Singapore.
distributed web server system load balancing algorithms using qualitative [23] J.M. Galloway, K.L. Smith, S.S. Vrbsky. 2011, October. Power aware load
parameters, VSRD-IJCSIT, Vol. 1 (8), 2011, 592-600. balancing for cloud computing. In Proceedings of the World Congress on
[9] B. Radojevic, M. Zagar ‘‘Analysis of issues with load balancing algorithms in Engineering and Computer Science (Vol. 1, pp. 19-21).
hosted (cloud) environments” In proceedings of 34th International Convention [24] S. Pandey, A.K. Upadhaya, C.K. Jha, Suboptimal mechanism for load balancing
on MIPRO,IEEE, pp. 23-27 May 2011. in cloud environment, 2017 International Conference on Algorithms,
[10] Meenakshi Sharma, Pankaj Sharma, Sandeep Sharma, ‘‘Efficient Load Balancing Methodology, Models and Applications in Emerging Technologies
Algorithm in VM Cloud Environment”, IJCST Vol.3, Issue 1, pp.439-441, Jan. (ICAMMAET), 2017.
March2012. [25] R. Aggarwal, L. Gupta, Load balancing in cloud computing, Int. J. Comput. Sci.
[11] T. Rao, K.T. Reddy, V.K. Sai Kiran, P. Thirumala Rao, B. 2015. Mitigation of Mobile Comput. 6 (6) (2017) 180–186.
insider attacks through multi-cloud, Int. J. Electric. Comput. Eng. Volume 5, [26] M. Bahrami, O. Bozorg-Haddad, Xuefeng Chu. 2018, Advanced Optimization by
Issue 1, 1 February 2015, Pages 136-141. Nature-Inspired Algorithms, PP.159-166, Springer.