International Journal of

INTELLIGENT SYSTEMS AND APPLICATIONS IN

ENGINEERING
ISSN:2147-67992147-6799 www.ijisae.org Original Research Paper

Adaptive Load Balancing in Cloud Computing Environment

1 2
* Santosh T. Waghmode , Bankat M. Patil

Submitted: 23/10/2022 Revised: 20/12/2022 Accepted: 04/01/2023

Abstract. The cloud computing platform load balancing is one of the challenge. Effective or efficient resource use in the cloud
environment also known as load balancing became an important concern with the cloud computing platform's rapid development in users
and their need for a variety of services. In order to enhance network capacity and reduce response time, load balancing play is important
role in storage system. Our primary goal is to provide a novel load balancing method that, by combining efficient scheduling and cloud -
based techniques, can balance incoming requests load from users across the world that are located in various regions to get data from
distributed data sources. In addition to effectively utilizing cloud resources and improving task processing times, also worked on
develop a dynamic load balancing algorithm that also ensures elasticity in a cloud environment. Divisible weighted round-robin, round-
robin, and weighted least connection alorithm are the primary load balancing algorithms discussed in this paper.

Keywords: Cloud computing, Divisible Weighted round robin, Least connection alorithm, Web server.

1. Introduction
Every day, the need for computing and storage resources
increases along with the IT industry. Large amounts of data are
produced and transmitted through networks, demanding an
increase in the need for computing resources. Load balancing
[2] [3] is a technique that divides the workload among several
nodes in the environment so that no nodes is overworked or left
idle for any span of time. A load balancing method that is
effective will ensure that each node in the system performs
approximately equal amounts of work. The Load balancing
algorithm objective is to distribute work to online services that
are sent to the cloud domain, enhancing the overall response
time and ensuring efficient resource use. In order to meet user
requests and increase system performance, load balancing in
the cloud primarily aims to dynamically balance the load across
the nodes[2]. An effective or optimal load balancing algorithm
helps to make the best use of the resources that are available,
ensuring that no node is overloaded or load balance.
Fig. 1 Diagram for Load balancing [25][31]
Scalability, the minimization of bottlenecks, and a shorter
The main drawbacks of existing systems include excessive
response time are all benefits of load balancing. It has been
power consumption from overloaded CPUs, increasing costs,
determined that effectively distributing the jobs throughout the
and long maximum execution times [12]. The primary goal of
system is an NP-complete problem [5].
this work is to focus on creating a framework for dynamic
virtual server allocation implementing a resources webserver in
the cloud. In order to achieve this goal, adequate resource
provisioning is required, along with dynamic virtual machine
1 placement that reduces total energy use [29] [31].
Dr. Babasaheb Ambedkar Marathwada University, Aurangabad
(MS) India
1*
[email protected] 1.1 Load Balancing in Cloud Computing Envoirment
2 The dynamic allocation of workload is a key issue in cloud
Dr. Babasaheb Ambedkar Marathwada University, Aurangabad
computing. The entire processing time needed by a machine to
(MS) India
2
[email protected] complete all of the tasks that have been delegated to it is
referred to as the machine's workload [1]. By distributing the
workload among the processors, load balancing works to
increase system performance. Increased resource efficiency
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 209
leads to improved overall performance and maximum client dynamic load balancing solutions in divisible weighted round
satisfaction, which are two advantages of workload. By evenly robin algorithm:
distributing the workload among all virtual machines in a cloud • By creating effective job allocation algorithms that
system, throughput is increased and response time is decreased assign each task to the best possible processor node for
[1] [3]. completion, the processing speed of divisible weighted round
1.2 Types of approaches to load balancing based on the robin may be fully realized.
cloud environment • Distributed systems are facing challenges that must
Static Load Balancing: In static algorithms [3], system be resolved in order to meet the growing demands for better
characteristics such as processor speed, memory capacity, performance and services for various distributed applications.
performance, and information on user needs are already known. Distributed systems are continuing to increase in scale,
These algorithms have important limitations in the event of a heterogeneity, and different networking technology.
sudden failure of a system resource, tasks, and the inability to • Jobs execute at different speeds on computer systems
relocate tasks while they are being executed for load balancing. nodes as a result of the heterogeneity of those nodes. Therefore,
Dynamic Load Balancing: Dynamic algorithms make load in this article research work focus on arranging i. e. divisible
balancing decisions based on the system's current state, weighted round robin in contexts.
therefore prior system knowledge is not necessary. The 1.5 Objectives of Load Balancing
limitations of a static method will be overcome by doing this. The objectives of load balancing are:
The dynamic algorithms are challenging, but they can offer • To significantly system improve performance
higher speed and fault tolerance [1] [25]. The dynamic load • To protect against system failures and maintain web-
balancing method uses some policies. These are best stated as site traffic
follows: • To Handle Sudden Traffic Burst
• Transfer Policy: Transfer policy refers to the decision • To keep the system stable.
made by the dynamic load balancing algorithm while transfer-
ring a job from a local node to a remote node. 2. Literature Review
• Information Policy: Information policy or infor- In the past few years, a large number of static and dynamic
mation strategy refers to the gathering of data about the system load balancing algorithms have been proposed and developed
node for the load balancing algorithm. for cloud environments. Static algorithms required the task and
• Load estimation Policy: This policy calculates the resource information at the time of build, i.e., before the
overall workload on a processor or device. execution never started. This kind of algorithm doesn't
• Process Transfer Policy: It is used to decide whether continuously monitor the state of the virtual machine [1].
a task should be carried out locally or remotely. However, because the cloud environment is dynamic and the
• Migration Limiting Policy: It provides a constraint on cloud service provider never predicts incoming task requests
how frequently a task may move from one machine to another. and the present status of all cloud resources at the time of
1.3 Types of Load Balancing Algorithms compilation, static algorithms are not appropriate for it. After a
Divisible weighted round robin: Depending on the standards predetermined period of time, dynamic load balancing
the network administrator specifies, each server is given a algorithms continuously monitor the VM [1] [2].
weight; the most typical criterion is the server's ability to The focus of this publication during analysis is dynamic load
handle traffic. The server receives more client requests the balance. Additionally, depending on the migration procedure,
greater the weight. dynamic load balancing is known as fixed and iterative load
Least connections method: In tis algorithm calculates the load balancing. While iterative load balancing requires multiple
distribution for each node based on the quantity of connections. steps, direct load balancing only needs one phase to balance the
The quantity decreases when a connection is terminated or load. Interactive dynamic load balancing is the method for
times out while it increases when a new connection is balancing the loads that is further divided into the diffusion and
established. directional exchange methods.
Dynamic algorithm: The dynamic method seeks for the The diffusion method diffuses more lead from fully loaded
network's lightest server first and recommends it for load nodes to lightweight loaded nodes [3] [4][12].
balancing. To boost system traffic, real-time communication For load balancing in cloud computing, there are various
with the network is necessary. heuristic and meta-heuristic dynamic algorithms available, but
Dynamic algorithms are capable of producing load transfer each algorithm has its own limitations, meaning that no single
decisions based on the system's current state. In order to algorithm can achieve all the goals at once because some
achieve this goal, adequate resource provisioning is required, parameters cannot be fully utilized or reduced simultaneously,
along with dynamic virtual machine placement that reduces such as cost and time [5][6].
total energy use. Execution time reduces as the cost of adding more virtual
1.4 Motivation and Research demands machines rises because more virtual machines are available to
Generally has the dynamic load balancing problem taken into do the task at hand. In the hybrid task scheduling algorithm, the
account the heterogeneity of the computing system in the equal distribution of tasks is the main goal in order to lower the
related study activity [1]. An effective mapping of jobs to the communication costs of cloud resources while also achieving
collection of computing nodes is the primary goal of the load balanced scheduling according to the computing capacity of
balancing problem in distributed computational environments each node. Created a load balancing strategy to reduce task
[3] [25]. response times in a cloud environment [6][8].
The following are the main motivations for researching
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 210
The distributed system has achieved a dynamic load balancing that application. When network and disc speeds produce
using two different techniques, such as centralized load variations in device performance when multiple users share
balancing and distributed load balancing. In the centralize load storage resources through striping [16] [17] [19].
balance technique, decisions on load balancing are made by a In this article broker-based architecture for task scheduling and
centralized model that manages the distributed system [8]. elasticity in cloud environment using cloud analyst as a tool for
Every node in a distributed system has metadata information implementation [19] [25]. The proposed work is restricted by
that is kept on one of the architectural control nodes. Metadata the load balance capacity to keep sessions across various
is used by a cloud - based data load balancer, and decisions virtual machines because it lacks the session affinity capability
about file allocation or file movement balance the cluster [9] [26]. novel method that calculates the computational time of
[10]. tasks while taking into account the importance of the work and
The author publishes one international journal, Load balancing the cost of the project.
in multi-agent distributed computer systems[11]. In this article,
the load balance is attained with the use of agents. After that, 3. Methods
agents are deployed throughout all nodes based on reputation Cloud load balancing is a technique for allocating workloads
value. According to the credit value, the migration node is and computing resources in a cloud computing environment.
selected. A method for distributed file system load balancing There are two elementary solutions to overcome the problem of
called adaptive data loading migration [12] [14] provides a overloading on the servers. The first is a single-server approach
comprehensive explanation of various types of load. in which the server is updated to a higher performance server.
This research presents the network, input and output, and The new server can also quickly become overloaded, calling
storage loads that are put on the system. Although network load for another update. The upgrade process is very expensive and
and input and output load increase, the loading of any disc time-consuming. The second approach uses many servers to
capability stays the same to the current data server Access build a scalable service system on a cluster of servers. Building
condition of multiple clients [15]. The solution is provided by a server cluster system for network services is therefore more
networked storage systems with paper users and centralized practical from a business viewpoint as well as more scalable.
information processing. In this method, migration decisions are
made by each user or application depending on what is best for

Fig. 2 Work load balancing system [25][30]

The technique was tested using a simulator. Cloud elements ware, and virtual machine -2 has a service pack for software
including Virtual Machines, Data Centers, Hosts, and Cloud and network resources.
servers are included in the simulator. Cost, disc size, RAM • Cloud Servers: The tasks or jobs that are now operat-
size, storage capacity, and other configuration details for the ing in the virtual machine are known as cloud servers. The
various components are stated[31]. An abstract cloud matching resource pack is processed in response to the client's
environment model is created by the integration of all the parts. request, and the requested service is given to the client.
Simulator Modules: • Load Balancer: A divisible weighted round robin ap-
• Virtual machines: Virtual machines make up each da- proach is used by default in the load balancing policy. Without
ta center and are responsible for providing the clients' required taking into account the load on each virtual computer, this
resources. rounds-robin distributes all incoming requests to the available
• Service Packs: There are service packs for each virtu- virtual machines.
al machine, e.g. virtual machine-l has a service pack for soft-

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 211
 Table 1: Adaptive Workload Balancer
Algorithm 1.1 Adaptive Workload Balancer

Input: Number of Jobs such as J1, J2, …Jn, (J –Jobs) on

VM M1,M2,M3.......Mn

Output: Jobs allotted on available VMs.

1. Start
2. Generate n task as jobs j1 , j2 ………..jn gets at the load balancer
where i=1,2……n.
3. Classified based on their size, order the jobs in decreasing order.
4. Initially, a load balancer would determine whether VMs were available.
5. Assign the job tasks t1, t2, t3… tn to VM servers as per weight based
server state.
6. IF (No. Of VM >= 1) then
1. C
arry out checks on the jobs, or the tasks.
2. Keep the virtual machines available (workstations).
3. The VMs should be arranged in order of their weight-based load.
4. The job would be allocated to an available VM through a specific data
centre after the load balancer searched the listed VMs' available VM IDs and Data
Center IDs.

7. Else
load balancer will look for weight based server
8. Once a job has finished running in a virtual machine, the related job ID
and the output data will be returned.
9. F
ind out how long it took to response to the results.
10. I
f the pending jobs are still available, then go to step 6.
11. End.

In out goal in designing this system was to produce a Master Slave Distribute. Information about files, blocks, file to
distributed system with at least five nodes that would deliver a block mapping, block to minion mapping, and block storage for
collection of web services. Load balancing is a strategy that Minion will all be kept in the file system namespace of the
seeks to enhance network performance as well as the output master. A file will be generated by the master and stored on
from database servers. To boost productivity and speed up blocks and a minion. The first minion will get it from the client
response times, distributions of visitor load balancing over a and pass it along to the next. The client will read the blocks in
multi-course community are implemented [1] [3]. order from a list of locations that the master sends for reading.
The network protocols respond to the operational request given If a minion's reading is unsuccessful, it will move on to the
by the client throughout the duration of the request, another minion
maximizing the space that is available and ensuring that no The round robin, weight round robin random & dynamic load
server is overworked while preserving performance.The balancer technique is used to create load balance modules in
Divisible weighted round robin load balancer distributes clients Python. Round robin load balancing is one of the simplest
to the servers that are still online if a server goes offline [12] techniques for dividing client requests across different servers.
[31]. To find out if the underlying server can manage the Each server in the group receives a client request as the round
amount of traffic, they do health checks. robin load balancer moves down the list of servers.
Performance measures of virtual machines in terms of capacity
4. Experiment & Result of virtual machine, execution of time for job, total load for all
The user makes a request to the Load Balancer. A Django virtual machine, throughput, scalability and latency as below:
application sends information to a Master slave server for the
Process of Jobs. Data sent across multiple Minions by the

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 212
 Fig. 3 adaptive workload balancer flowchart

Using the given equation, we compute each virtual machine's

capacity and the total capacity of all virtual machines: Assign load at the virtual machine is:
𝐶𝑎𝑝𝑎𝑐𝑖𝑡𝑦 𝑜𝑓 𝑉𝑀𝑠 = 𝑃 ∗ 𝑁 + 𝑠𝑖𝑧𝑒
𝐿𝑜𝑎𝑑 𝑜𝑓 𝑉𝑀𝑠 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑗𝑜𝑏 ∗ 𝐿𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝐽𝑜𝑏
𝑃∗𝑁
𝐵𝑊𝑖𝑗 (1)
(4)

Where P = Processing of Speed,

The throughput statistic is determined by taking the entire
N=Number of CPU is used,
work, translating it to hours from seconds, scaling it by the
Bi-j=bandwidth between VM
database volume (SF), and dividing it by the total amount of
Using the given equation, to calculate how long it will take a
time (Ts) that has passed between the commencement of the
job to execute on a virtual machine:
first query and the completion of the last query or update
𝐸𝑥𝑒𝑐𝑢𝑡𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 𝑓𝑜𝑟 𝑒𝑎𝑐ℎ 𝑗𝑜𝑏 = 𝐽𝑜𝑏 𝑠𝑖𝑧𝑒/𝑃 ∗ 𝑁
function.
(2)
The frequency of successfully sent queries across various
servers is known as throughput. The quantity of service
Using the given equation, to calculate total load for all virtual
delivered by the system in a single time unit is known as
machines i. e datacenter:
throughput. The ratio is calculated by formula such as:

𝑇𝑜𝑡𝑎𝑙 𝐿𝑜𝑎𝑑 = ∑𝑚
𝑗=1 𝐿𝑜𝑎𝑑 𝑜𝑓 𝑉𝑀𝑠 (3)
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 213
 𝑦 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐽𝑜𝑏𝑠 /𝑝𝑒𝑟𝑖𝑜𝑑 𝑜𝑓 𝑒𝑣𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛 𝐿 = 𝑓(𝐿𝑐, 𝐿𝑠)
(5)

Scalability is the capacity of a process to grow to handle that Step 5 in algorithm, determine utilization virtual machine
growth as well as its ability to convey load balancing from using,
clients to servers in a network. The transmission load ratio
(TLR) is determined using a formula such as: 𝑈 = ∑𝑀𝑖
𝑗=1(𝑈𝑖 … . . 𝑗) (7)
𝐴𝑠𝑠𝑖𝑔𝑛𝑒𝑑 𝐿𝑜𝑎𝑑
𝑇𝐿𝑅 = Determine, mean of utilization of VM is
𝑂𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛 𝑑𝑢𝑟𝑎𝑡𝑖𝑜𝑛
(6)
𝑝 1
𝑀 = ∑𝑘=1(𝑋𝑘) (8)
𝑃
For distributed web servers, load-balancing (salb), a file-by-file Where Xk is required total utilization of VMs in time i.e. k
device, enables load-balancing. The mathematical approach for P is denoted total time required
load balance in CC as shown in: Reduce the system's load
imbalance and the total transfer costs incurred by the system as Variance of utilization of VMs is
a result of load balancing procedures. 𝑉=
1
∑𝑝𝑘=1(𝑋𝑘 − 𝑀)2 (9)
𝐿𝑠 = 𝑓𝑠𝑖𝑛𝑖 𝑝
Standard deviation is
𝐿𝑐 = 𝑁𝑜. 𝑜𝑓 𝑟𝑒𝑞𝑢𝑒𝑠𝑡 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑡𝑖𝑚𝑒 𝑆𝑡𝑑𝐷𝑒𝑣 = √𝑉 (10)

Load considered in SALB is Average of jobs at VMs is

𝐴𝑣𝑔 = 𝑈 + 𝑀 ∗ 𝑆𝑡𝑑𝐷𝑒𝑣 (11)

Fig. 4 Throughput over time result

In figure 4 shows Throughput is a metric measure how many algorithms. On the X-axis of this graph, time has passed. The
data packets are successfully sent each second and is used to elapsed time in millisecond may be either actual or relative
evaluate performance. The results indicate that the proposed time. Y-axis represents to show how much data was transferred
method provides higher throughput than the conventional from the server to the client (in bytes/KB/MB) [3] [25].

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 214
 Fig. 5 Latency (ms) over request result of algorithms

Fig. 6 Comparison of response time using proposed algorithms

In figure 6 compares the response times of the least connection times are shown in fig 6. In comparison to round robin and
algorithm, load balancing, and the divisible weighted round least connection algorithm, the time required to conduct the
robin algorithms. Figure 6 provides the execution costs for the load balancing with divisible weighted round robin algorithm is
three algorithms. The three algorithms' execution response reduced by half [3] [25].

Fig. 7 Comparison of execution cost using proposed different algorithms

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 215
In figure 7 show bar plot for adaptive workload balance model between proposed algorithms, red color indicate for divisible
with the help of different load balancer algorithms, the weighted round robin and green color for least connection
proposed algorithm successfully distributes incoming algorithm and blue color for eastic load balancer algorithm
workloads across several virtual machines and processing indicated [3] [25].
speeds in various data centers. Comparison of execution cost

Table 2: Comparatıve Analysıs of Algorıthms [25]

Parameters DWRR RR LCA

Scalability Yes Yes Yes
Execution Time Fast( 8 Sec) Fast Slow
Turnaround time Less Less More
Throughput Low Low Low
VMs Load Distribution Fast Slow Slow

Table II compared the proposed algorithm with the existing reduce response time and complete out load balancing without
algorithms. Divisible weighted Round Robin, round robin, and producing any delays, a divisible weighted round robin load
least connection method are the algorithms that are come up for balancer has been implemented.
comparison.
Experiments are run, data are assessed, and discussions are References
presented on three metrics: average success rate, resource [1] A. Nuaimi, N. Mohamed, A. Jaroodi, “A survey of load
scheduling efficiency, and response time. This demonstrates balancing in cloud computing: challenges and algorithm”,
the effectiveness of the divisible weighted round robin in Proc - IEEE 2nd Symp. Netw. Cloud Comput. Appl
algorithm. .NCCA, 2012, pp. 137–142.
Impact of Average Success Rate: Average Success Rate (ASR) [2] Ammar Rayes, Bechir Hamdaoui, Mehiar Dabbagh and
measures how many requests from users are successfully Mohsen Guizaniy, “Energy-Efficient Resource Alloca-
handled by the virtual machine grid through the resource tion and Provisioning Framework for Cloud Data Cen-
manager at any given moment in a cloud environment. The ters”, IEEE 2015.
following mathematical formula represents the average success [3] Aditya Narhe, Santosh Waghmode, 2019. SQL Query
rate. Formation for Database System using NLP. International
𝑉𝑖𝑟𝑡𝑢𝑎𝑙 𝑚𝑎𝑐ℎ𝑖𝑛𝑒 𝑔𝑟𝑖𝑑 ′ 𝑠 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑟𝑎𝑡𝑒 𝑡𝑜 𝑢𝑠𝑒𝑟 𝑟𝑒𝑞𝑢𝑒𝑠𝑡 Journal of Engineering Research and, V8(12).
𝐴𝑆𝑅 = ∗
Total no.of user requests [4] Babu D, Venkata P. Honey bee behavior inspired load
100 (12) balancing of tasks in cloud computing environments.
Impact of Resource Scheduling Efficiency: The ratio of the Appl Soft Comput. 2013; 13(5):2292–2303.
amount of resources allocated based on user request to the total [5] D. Fernández-Baca: Allocating modules to processors in
number of resources in the cloud is referred as the resource a distributed system, IEEE Transactions on Software En-
scheduling efficiency. Efficiency in resource scheduling can be gineering, Vol. 15, No. 11, pp. 1427-1436 (1989).
defined mathematically as follows. [6] Elzeki OM, Reshad MZ, Elsoud MA. Improved maxmin
Resource Scheduling Efficiency = algorithm in cloud computing. Int J Comput Appl Tech-
No.of resources are scheduled nol. 2020; 50(12):22–27.
∗ 100 (13)
Total no.of resources
[7] Ektemal Al-Rayis, Heba Kurdi, “Performance Analysis
of Load Balancing Architectures in Cloud Computing”
Impact of Response Time: Response time is the amount of time
IEEE Modeling Symposium (EMS), 2013 European, 20-
needed by a distributed divisible weighted round robin
22 Nov. 2013, 978-1-4799-2577-3.
algorithm to respond while scheduling cloud resources. The
[8] Haozheng Ren, Yihua Lan, Chao Yin,” The Load Balanc-
following is a mathematical formulation of response time.
ing Algorithm in Cloud Computing Environment”, 2012
𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑡𝑖𝑚𝑒 = 𝑛(𝑢𝑠𝑒𝑟𝑠 𝑟𝑒𝑞𝑢𝑒𝑠𝑡) ∗
2nd International Conference on Computer Science and
𝑡𝑖𝑚𝑒(𝑟𝑒𝑠𝑝𝑜𝑛𝑒𝑑 𝑟𝑒𝑞𝑢𝑖𝑟𝑒𝑑) (14)
Network Technology, IEEE 978-1-4673-2964-4/12.
5. Conclusion [9] Kumar, M, Subhash CS. Priority Aware Longest Job First
The proposed divisible weighted round robin in this paper (PA-LJF) algorithm for utilization of the resource in
successfully handled the incoming jobs for distribution among cloud environment. 3rd International Conference on
various virtual machines located in various Data Centers [12]. Computing for Sustainable Global Development (IN-
In order to meet the growing demand for high performance web DIACom). New Delhi: IEEE; 2016.
servers produced by the introduction of improved web [10] Load Balancing Algorithm Based on Estimating Finish
environments, cluster-based web servers are used. Through the Time of Services in Cloud Computing Nguyen Khac
results of experiments, we evaluate the effectiveness of the Chien, Nguyen Hong Son, Ho Dac Loc, 2016 18th Inter-
proposed load-balancing technique [12]. According to Round- national Conference on Advance Communication Tech-
robin scheduling, divisible weight Round-robin scheduling, and nology (ICACT).
weighted least connection alorithm scheduling the approach [11] Maha A Metawei, Salma A. Ghoneim Sahar M Haggag,
greatly reduces both the load range and load variance. To Salwa M Nassar, Load Balancing in distributed multia-
gent computing system, Elsevier Aims Shams Engineer-
International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 216
 ing journal 2012. and Information Technology, 2010, pp. 108–113.
[12] Mohit Kumar & S. C. Sharma “Dynamic load balancing [22] S. Sonkar and M. Kharat, “A Review on Resource Allo-
algorithm to minimize the makespan time and utilize the cation and VM Scheduling Techniques and a Model for
resources effectively in cloud environment” International Efficient Resource Management in Cloud Computing
Journal of Computers and Applications, 23 Nov 2017. Environment,” in ICTBIG, Indore, India, Nov. 18-19,
[13] M. Randles, D. Lamb, T. Bendiab, “A comparative study 2016, pp. 1-7, doi: 10.1109/ICTBIG.2016.7892646.
into distributed load balancing algorithms for cloud com- [23] S. Abed, D. S. Shubair, "Enhancement of task scheduling
puting”, presented at Int. Conf. in Adv Inf Netw Appl technique of big data cloud computing", presented at Ad-
Work, pp. 551–556. vances in Big Data Comput. and Data Commun. Syst.,
[14] N. Ajith Singh, M. Hemalatha, “An approach on semi 2018, pp. 1-6.
distributed load balancing algorithm for cloud computing [24] Shridhar G.Damanal and G. Ram Mahana Reddy, “Opti-
systems” International Journal of Computer Applications mal Load Balancing in Cloud Computing By Efficient
Vol-56 No.12 2019. Utilization of Virtual Machines”, IEEE 978-1-4799-
[15] Naha RK, Othman M. Brokering and load-balancing 3635-9/14.
mechanism in the cloud revisited. IETE Tech, Rev., [25] Waghmode, S. and Patil, B., 2021. Load Balancing
2014; 31(4):271–276. Technique in Distributed Systems: A Review. In: 2021,
[16] O. Kaneria, R. Banyal, "Analysis and improvement of 2nd Global Conference for Advancement in Technology
load balancing in cloud computing", presented at Int. (GCAT). Bangalore, India: IEEE.
Conf. on ICT in Business Industry & Government, 2017. [26] Waghmode, S. and Patil, B., 2022. Optimized and adap-
[17] Onur Destanoglu, F. Erdogan Sevilgen, Randomized Hy- tive dynamic load balancing in distributed database serv-
drodynamic Load Balancing Approach, IEEE 1530-2016, er. [Online] Digital-library.theiet.org. Available at:
2008. <https://digital-
[18] Shanti Swaroop moharana, Rajadeepan d. Ramesh & li-
Digamber Powar ,” Analysis of load balancers in cloud brary.theiet.org/content/conferences/10.1049/icp.2022.06
computing” International Journal of Computer Science 07>, 2 October 2022.
and Engineering (IJCSE)ISSN 2278-9960 Vol. 2, Issue 2, [27] X. Shao, M. Jibiki, Y. Teranishi, N. Nishinaga , “Effec-
May 2017, 101-108. tive load balancing mechanism for heterogeneous range
[19] Sidra Aslam Munam, Ali Shah “Load Balancing Algo- queriable cloud storage”, presented at IEEE 7th Int. Conf.
rithms in Cloud Computing: A Survey of Modern Tech- Cloud Computing Technology, 2015, pp. 405–412.
niques” 2015 National Software Engineering Confer- [28] X. Shao, M. Jibiki, Y. Teranishi, N. Nishinaga , “Effec-
ence (NSEC 2015),IEEE, pp:30-35. tive load balancing mechanism for heterogeneous range
[20] S. Abed, D. S. Shubair, "Enhancement of task scheduling queriable cloud storage”, presented at IEEE 7th Int.
technique of big data cloud computing", presented at Ad- Conf. Cloud Computing Technology, 2015, pp. 405–41.
vances in Big Data Comput. and Data Commun. Syst., [29] Zhipeng Tan, Wei Zhou, Dan Feng, and Wenhua Zhang ,
2018, pp. 1-6. ALDM: Adaptive Loading Data Migration in Distributed
[21] S. Wang, K.Q. Yan, W. P. Liao, SS Wang, “Towards a File Systems, IEEE transactions on magnetics, vol. 49,
load balancing in a three-level cloud computing net- no. 6, June 2013.
work”, in Proc. 3rd IEEE Int. Conf. on Comput. Science [30] https://www.znetlive.com/multi-cloud/

International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2023, 11(1s), 209–217 | 217