Exploring Load Balancing: Issues, Methods and Strategic Outlook in

Cloud Computing

Zainab Khan and Kavita Agrawal

Department of Computer Science and Engineering, Integral University, Lucknow, India

Keywords: Cloud Computing (CC), Load-Balancing (LB), Resource Allocation, Heuristics Algorithms Approach, ACO,

PSO, Blockchain.

Abstract: Through the provision of scalable, on-demand resources via the Internet, cloud computing has completely

transformed contemporary computing. Effective load-balancing remains a critical challenge in achieving

efficient resource use, reducing response times, and preventing system overload. A comprehensive evaluation

of LB strategies in distributed computing is given in this document, which divides them into conventional and

contemporary categories, such as static, dynamic, heuristic, and AI-based approaches. It also examines

important security, fault tolerance, and energy efficiency issues. And scalability. The study highlights new

developments influencing load balancing in cloud environments going forward, such as edge computing,

blockchain integration, and machine learning-driven optimization. The purpose of this survey is to give

researchers and practitioners useful information about how load balancing is changing, enabling

improvements in the effectiveness and performance of CC.

1 INTRODUCTION

The use of CC has become a major craze in the last

few years, leading to significant advancements in

distributed systems and the development of extensive

computer networks. Cloud services are provided to

customers worldwide by CC companies like IBM,

Amazon, and Google. Under this new paradigm, end

users can access apps and services whenever they

want rather than having to install them on their local

computers (M. Shahid, et.al, 2020)

The foundation and key component of cloud-

based applications is virtualization. Inefficient

handling of the migration process and the allocations

of VMs can greatly impact the way on-demand as

well as scalable services are rendered to customers

(D. Shafiq,et.al, 2021) Among the top three

difficulties with Cloud computing, according to (M.

Shahid, et.al, 2020), mentioned as cloud performance.

This study intends to improve the infrastructure as a

service model’s resource allocation, a key concept

idea in cloud computing, by balancing the resources

offered to the customers with the amount of work as

well as requests made by users on servers (D.

Shafiq,et.al, 2021), Allocating resources is one of the

challenges with CC, and it also plays a role in LB.

This problem is also present in systems for wireless

communication, it’s essential to allocate resources in

a fair and balanced manner while also considering

user priorities (F. Zabini,et.al.2017)

There are two ways to classify cloud computing:

by location or by services provided. A cloud has the

potential to be categorized as public, private, hybrid,

or community, depending on its location (J. Shah,

2017), Anybody can use public cloud services, and

the infrastructure is housed on a service provider’s

property. Public clouds are the most economical, but

they are also the most susceptible to different types of

attacks. Access to a private cloud is limited to a single

person or entity. Although it costs more, it offers the

user the highest level of security and control.

Combining both public and private clouds for various

uses depending on organizational needs is known as

a hybrid cloud. A community cloud is made up of a

shared infrastructure that is utilized by numerous

organizations with similar management and data. (P.

Kumar,2019)

The following are the study’s primary goals:

• To investigate different load-balancing

strategies that are discussed within the

literature.

650

Khan, Z. and Agrawal, K.

Exploring Load Balancing: Issues, Methods and Strategic Outlook in Cloud Computing.

DOI: 10.5220/0013941100004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 5, pages

650-662

ISBN: 978-989-758-777-1

• To categorize different LB methods as

well as offer a summary of the difficulties

and problems that LB currently faces.

• To provide an overview of potential study

topics for future load-balancing technique

improvements.

Challenges of Cloud Computing.

Multiple challenges exist in the usage of CC. Three

important challenges in CC include data protection

(A. Chaturvedi,et.al, 2019) data availability, security,

management of execution and load balancing, and

fault tolerance (R. Khan and M. Ahmad, 2016),

Protection of data: One important factor that must be

considered is data protection. The privacy issue

persists when data is stored on the cloud. Likewise, the

organization’s privacy concerns are exacerbated by

the fact that the exact location of repository sites is

frequently unknown. Most current models use data

centers to protect knowledge through firewalls (A.

Chaturvedi,et.al, 2019).

1. Problems with data availability and retrieval:

SLAs will fully adhere to the business's

needs. Here, the operational staff is crucial to

managing system time and supervising SLAs

2. Security: To safeguard data and the

virtualized Internet, apps, services, as the

associated CC infrastructure, security of

cloud computing, more popularly, cloud

protection, includes a wide array of laws,

technologies, applications, and control

mechanisms. It falls under the subdomains of

information security, internet security, and

more generally, computer security (R. Khan

and M. Ahmad, 2016).

3. Managing execution: Each stack’s cores are

subject to tension adjustment. Additionally,

this boosts device output. Many recent

figures provide an effective use of resources

and a change to the stack. Cloud stacks can

be created in a variety of ways, including

memory, CPU, and structure stacks. The path

to center point overload and subsequent store

relocation to other centers is to alter the

strain.

4. Load balancing: One of the main problems

with the CC at the moment is load balancing,

which prevents some nodes from being

underutilized while others are overloaded, so

the idle ones need to be put to work. Price,

reaction time, dependability, effectiveness,

and use of resources are some of the QoS

metrics that load balancing may enhance (R.

Khan and M. Ahmad, 2016).

5. Fault tolerance: Because resource dropping

affects unit performance, job outcomes,

productivity, reaction time, and high quality,

FT is among the most important parameters.

Consequently, in order to identify errors, fix

them, and improve performance metrics, a

fault tolerance strategy is needed. To

guarantee the continuity of the essential

services and the program’s completion, fault

tolerance is a crucial consideration.

2 LOAD-BALANCING

One crucial technique for distributing workloads

among cloud computing users is load-balancing, such

as several computer resources to maximize resource

use, improve performance, and guarantee system

dependability.

Research environments rely on load balancing as

their essential practice to accomplish tasks that

combine artificial intelligence processing with large-

scale simulation, big data analytics, and high-

performance computing, which is essential for

handling large volumes of data.

The primary objective of LB is to efficiently divide

the workloads across different cloud endpoints,

preventing any node from being overloaded and

sometimes underloaded (J. Shah, 2017) To optimize

its utilization of resources and enhance the overall

reaction time, LB can be defined as the procedure for

allocating a load among network links on several

gadgets or groups of systems. It prevents excessive

asset replication and shortens the device's overall

waiting time. To distribute and process data without

waiting, requests are dispersed throughout servers

during this process. By shifting the device burden, LB

maximizes system performance (M. A. Hossain and S.

Roy,2019)

2.1 Characteristics of Load Balancing

Distributed network traffic or computational

workloads represent the technique of allocating

workloads among several servers to keep any one

server from becoming overloaded, which is referred to

as load-balancing. The following are the main features

of load balancing firewalls (A. Chaturvedi,et.al,

2019).

Evenly Traffic Distribution: To avoid bottlenecks

and guarantee peak performance, incoming requests

Exploring Load Balancing: Issues, Methods and Strategic Outlook in Cloud Computing

651

or workloads are effectively distributed among

available resources.

1) Excellent Availability and Dependability:

The system offers superb reliability

alongside dependability because it redirects

system traffic to functional servers to

maintain service accessibility when servers

fail (A. Jain and R. Kumar, 2016)

2) LB Scalability: The scalability parameter of

LB allows administrators to dynamically

add or remove servers for traffic demand

variations.

3) Failover and Fault Tolerance: The system

ensures service continuity through the

detection of broken elements by remedying

traffic diversion to functional servers.

4) Conservation persistent (Sticky Session):

The system needs sticky sessions to

maintain user requests within a single server

throughout their connection. IP hashing and

cookies, together with other methods,

operate at load-balancing to achieve this

goal (S. Afzal and G. Kavitha,2019).

2.2 Challenges of Cloud Computing’s

Load Balancing

LB is among the most pressing issues that require

particular attention out of all the difficulties that cloud

computing faces. This covers topics like VM security

and migration; user comfort with QoS and source

usage are given equivalent weight when looking for a

better way to increase cloud-based resource

utilization. An inventory of some load-balancing

problems is provided beneath. Table 1 shows the The

Load-Balancing Challenges Overview Is Presented.

1) Distributed Location-Based Nodes: To

compute at different places, cloud data

centers are usually dispersed. These centers

use a centralized network of dynamically

distributed nodes to process consumer

requests efficiently. There are several load-

balancing techniques with a narrow scope

that ignore factors like network and

communication latency, the distance between

distributed computing nodes, customer space,

and resource availability. It is difficult to

operate nodes in extremely remote locations

since more algorithms are not appropriate for

this setting (P. Kumar,2019)

2) A single failure instance: Certain algorithms

used for load-balancing are put forth in the

literature in situations where the centralized

node makes load-balancing decisions rather

than decision-making being divided among

several nodes. The entire computer system

will be impacted if the main components fail.

3) VM Mobility: Several virtual computers can

be constructed on a single physical unit,

thanks to virtualization. These VMs are

autonomous in their architecture and have

many configurations. It is suitable to move all

virtual machines (VMs) to a distant site using

the LB approach if a physical device is

overwhelmed.

4) Hypothesis for Perception: The authors are

making load balancing in the cloud a

homogenous node in the original question. A

switch that is dynamic is needed by CC

consumers, whose execution needs to be done

on heterogeneous nodes to have the most

effective network and reduce the response

time.

5) Data Handling: Old conventional storage

devices, especially hard disks, always

required massive resources and equipment

costs for hardware; this CC addressed.

Consumers can keep the data safely and

evenly with the help of the cloud without any

control problems. Storage is ever-growing

and demands in turn, redundancy of stored

data to maintain access and data availability.

6) Scalability: With cloud services of on-

demand scalability accessibility, people have

the opportunity to access resources to

downscale and downscale rapidly at any time

or scale up. A good load balance must adapt

quickly to variability in the computational

environment in terms of, for example,

changing requirements and conditions,

memory and device topology, and so on.

7) The Intricacy of the Algorithm: Algorithms

for cloud computing ought to be fast and

simplistic to accomplish. A stronger analysis

technique seeks to decrease the efficiency of

cloud systems and to excellence.

8) On-demand Self-Service: Among the most

important features that are related to CC is

scalability; materials could have

unsupervised provisioned or disseminated.

So, how do we apply or disassociate cloud

computing and its offerings, retaining

likewise efficiency like traditional systems as

well as the greatest resource (R. Khan and M.

Ahmad,2018)?

9) Control of Energy: The economy of the scale

is the advantage of cloud usage in energy

management. In the final analysis, saving on

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

652

power is the crucial element that renders a

worldwide economy possible, in which

limited companies will contribute to the pool

of international capital rather than separately

supplying their utilities (R. Khan and M.

Ahmad,2018).

Table 1: The Load-Balancing Challenges Overview is Presented.

Reference Title Challenges Description

P. and

R. Kumar

(2019)

A survey of the problems and

difficulties with cloud computing LB

techniques

Distributed

Geographical Nodes

Usually, cloud data centers

are dispersed to accommodate

computing in various places.

Single Instance of Failure

The centralized node makes

LB decisions; decision-

making is not dispersed

among several nodes.

VM Mobility

Building several virtual

machines on a single physical

unit is made possible by

virtualization.

Hypothesis for

Perception

For a network to be effective

and response times to be

shortened, CC consumers

require a dynamic switch

Rafiqul and M.

Zaman Oqail

A survey of LB difficulties in CC

On-Demand

Service

Flexibility is cloud

computing’s primary

characteristic; resources can

be distributed or delegated

automatically.

Energy Management

The economies of sales are

the advantages of energy

management, which promotes

cloud use.

2.3 Parameters of Load Balancing

In this work, some performance parameters are added

to the existing load balance parameters. Additionally,

new parameters may be added to the classification

based on their attributes if they are discovered in the

future. The cloud load-balancing parameters are

divided into four sets by taxonomy (R. Kaur and N.

Ghumman,2017). Table 2 shows the An Overview of

The Load-Balancing Parameters.

1) Load balancing metric with dependent

nature and qualitative characteristics.

2) Load balancing metrics that are independent

and have qualitative characteristics.

3) Load balancing metrics that are dependent

and have quantitative characteristics.

4) Load-balancing metrics that are independent

and have quantitative characteristic

Exploring Load Balancing: Issues, Methods and Strategic Outlook in Cloud Computing

653

Table 2: An Overview of the Load-Balancing Parameters.

References Title Parameters Descriptions

A. Jain & R.Kumar

(2016)

A cloud environment multi-

stage load-balancing method

Fault Tolerance

The algorithm’s capacity

to handle error situations

and its resilience to

failure.

R. Kaur & N. S.

Ghumman (2018)

Task-Based LB through

effective cloud computing

VM use

Reaction Time

It is determined by

deducting the completion

duration of an assignment

from the task delivery

eginning time.

Roy Sumendu (2019)

Performance evaluation od

load balancing algorithms

Overhead

Support the additional

expense of algorithm

inte

ration.

S. Afzal &

K. Ganesh

(2019)

A Systematic Review of a

Taxonomic Classification of

Load Balancing Metrics

Scalability

Possible for a device to

carry out consumer

operations within the

intricate traffic flow..

Improves system

consistency for the

efficiency parameter.

Performance

3 TECHNIQUES FOR

BALANCING LOAD

In cloud computing, load-balancing strategies can be

broadly divided into three categories: hybrid,

dynamic, and static. These methods aid in effectively

allocating workloads among several servers in order

to optimize resource use, enhance performance, as

well as guarantee high availability.

3.1 Static Techniques

Static load-balancing strategies don’t depend on the

state of the network; instead, they employ a static set

of rules. This approach requires specialized familiarity

with resources, including time of contact, node storage

and space, node capabilities for processing, etc., and

is not scalable. Although this method is fast and

effective, it usually fails to locate the connected

servers, which results in uneven resource allocation.

The main issue with this type of approach is that

decision-making does not give enough thought to the

system’s actual state. Therefore, distributed systems

cannot tolerate a state of constant change.

The techniques for static load-balancing methods

include:

• The Round Robin (RR) Method

• Minimum-Maximum (MM) Method

• Shortest Job First (SJF) Algorithm

3.2 Dynamic Techniques

The method starts its decision-making process after

reviewing the present system's situation. Each of these

tactics provides the benefits of shifting work from

machines with heavy loads to machines with light

loads (P. Kumar,2019)

The specified dynamic load-balancing methods

include:

• Least Connection Algorithm

• Throttled Algorithm

• Weighted Round Robin

4 EXAMINATION OF

COMPARISON IN LOAD-

BALANCING APPROACHES

4.1 Load Balancing Using

Round-Robin Algorithm

To handle multiple things happening at the same time,

the CPU uses a method called Round Robin. It is like

giving each task a tiny time slot and then just going

through the list again and again, ensuring everything

gets a chance to run. This section describes the

algorithm’s idea and how scholars have proposed

utilizing it to resolve CC load-balancing concerns.

Table 3 shows the Comparison Of The Enhanced

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

654

Simulation Tool Based On The Rr Algorithm’s

Performance.

1) Tailong and Dimri(2016) suggested that to

change CloudAnalyst’s current

responsiveness of the service broker

guidelines, the authors should use an

Optimize Response Time Algorithm was

modified (MORT). The method determines

the scheduling procedure after calculating

the waiting and response times for each

process. Although it can decrease response

time, for a dynamic cloud environment, the

method is less suitable because this method

didn’t solve the time quantum issues in RR.

2) Issawi et al. (2015), the author suggested a

Modified Optimize Response Researchers

have enhanced the quality of service (QoS)

in cloud-based apps by considering the

problem of burstiness in workload. An

abrupt increase in cloud service users results

in a load-balancing problem; therefore, both

situations should be considered in cloud

computing. To efficiently assign the tasks

you’ve been given to heavy loads of work on

VMs, an adaptive LB method

(RR+Random) is suggested. This algorithm

alternates between random and RR project

scheduling strategies.

3) Pasha et al. (2014), an RR-based task-based

LB approach is provided. The recommended

approach is an enhanced round-robin that

saves the most recent entry submitted by a

used base in a hash map, which decreases the

total reaction time in cloud applications (S.

A. Salman and M. K. Ahmed,2021)

Table 3: Comparison of the Enhanced Simulation Tool Based on the RR Algorithm’s Performance.

Author Parameters Tool Response Time Processing Time

Dimri &

Tailong (2016)

Processing and

response times

Cloud Analyst 151.72 ms 1.66 ms

Pasha &

Associate,

(2014

Processing time and

response time

CloudSim 299.91 ms 1.210 ms

Issawi &

Associate,

(2015)

Reaction time &

Processing time

NA 369.6 ms 314.9 ms

4.2 Load Balancing Using a Throttled

Algorithm

A load-balancing method known as the throttled

algorithm distributes requests among virtual machines

(VMs) according to their availability. Restricting the

number of active requests per virtual machine

guarantees effective resource use. This section

describes the Throttled algorithm’s idea, and in order

to solve load-balancing problems in CC, scholars have

proposed using it.

1) Phi and associates, 2018, suggested the TMA

algorithm, which tends to keep two virtual

machine (VM) tables with accessible and

hectic designations for each VM equitable

workload sharing. In contrast to the

conventional Algorithm, which keeps a single

database for each VM, making it more

challenging to find out if a Virtual Machine

(VM) is available whether or not. Reaction

Time decreased somewhat from 402.66 to

402.63 (ms) due to the algorithm. (A. Aliyu and

P. Souley,2019)

2) Banerjee and Ghosh, 2016, suggested that

according to the priority method, utilizing the

PMTA, or Modified Throttled Algorithm, has a

sufficiently quicker completion time above the

existing method. To perform top-priority jobs

at first, as well as to divide the workload evenly

between multiple virtual machines, it centers

on distributing tasks that arrive via a

preemptive queue to halt processes with lower

tiers from executing. Although the approach

decreased reaction and waiting times in

comparison to the existing TA as well as the

round-robin algorithm, it could nevertheless

lead to excessive reaction times as well as

craving for less important things.

3) Souley and Aliyu, 2019, suggested for

equitable workload distribution, TA, as well as

Exploring Load Balancing: Issues, Methods and Strategic Outlook in Cloud Computing

655

ESCE, the Hybrid Method is suggested,

maintaining a setpoint as each virtual

machine’s (VMs) top priority. In addition to

lowering response time, it is also reasonably

priced (S. Lian, et.al, 2017) Table 4 shows the

Comparison of The Efficiency of a Controlled

Algorithm-Based Simulation Tool.

Table 4: Comparison of the Efficiency of a Controlled Algorithm-Based Simulation Tool.

Algorithm Tools Response Time Processing Time

TMA CloudAnalyst Tool 402.6 ms 173.0 ms

Priority-based altered throttled algorithm

(PMTA)

CloudSim

Tool

0.06 ms 1.32 ms

Evaluation of a hybrid approach’s performance

(TA+ESCE)

--------- 315.68 ms 7.57 ms

4.3 Load Balancing Using Machine

Learning

The goal of machine learning (ML), a branch of AI, is

to teach systems to carry out novel tasks without

explicit programming. Training is the process of using

models using statistical techniques based on the past

data to produce models that can predict previously

unknown values. By guaranteeing service quality and

adherence to set SLAs, an intelligent load balancer

gives cloud providers a competitive edge. The

intelligent models listed below were examined. Table

5 shows the An Overview Table Displaying The Ml

Load-Balancing Strategies That Have Been

Reviewed.

1) S. Liang et al. suggested a load balancer to

manage traffic in the data center. The

quantification of load traffic was predicted

using a Bayesian network, which was then

integrated with reinforcement learning to

incorporate a self-adjustment criterion and

decide on the appropriate course of action.

The method used included delocalizing

processing.

2) J. Kumar and A. K. Singh created this way to

foretell the amount of work in the data center

on the cloud. This approach blends ANN with

self-adaptive differential evolution (SaDE).

User requests were combined into the periods

that functioned as historical information. The

ANN was trained using back data and actual

workload components. Future work in the

data center was predicted using the resulting

model. NASA and Saskatchewan server

datasets were used to train the model.

3) A. Kaur et al. employed regression using

deep learning to forecast the ongoing task

timeline based on computing time and cost.

Three concealed levels of convolution neural

networks, the layer of activation composed of

the ReLU function, and a pooling layer were

all intended components of the network for

profound learning. The budget, as well as

schedule parameter data from larger

operations, made up the training data.

4) A. Abbas, D. Sutter, and S. Worner employed

QNN as a type of neural network built on the

ideas of quantum computing. It’s been

discovered that quantum circuits operate

similarly to ANN. The workload that the

cloudlets would produce was predicted using

the QNN model. To modify the qubit network

weights, their prototype employed the

activation function is the CNOT gate in both

the output as well as hidden layers.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

656

Table 5: An Overview Table Displaying the ML Load-Balancing Strategies that have been Reviewed.

Title Model Parameters / Data Used LB-Problems

An SDN controller

based on

reinforcement

learning

The Bayesian network &

reinforcement learning

Data on network traffic

Security and stability of

networks

Workload prediction

ANN & self-adaptive

differential

Customer requests are sent

to time units

Workload distribution

Deep learning CNN, Regression Job flow information Qos, resource utilization

Quantum-based Load

Balancing

Evolutional QNN

(EQNN)

Workload logs for cloudlets

Distributed resource

scaling

4.4 Ant Colony Optimization for Load

Balancing

ACO is among the effective optimization techniques

for resolving LB issues in CC. ACO’s algorithm

imitates the behavioural patterns of a colony of Ants

looking for the second-best path between their nest as

well as the source of food. method eventually

converges to an ideal answer when the shortest path is

strengthened over time by a greater concentration of

pheromones. Some of the reviewed research papers

are mentioned below. Table 6 shows the Examination

of Ac0 Algorithm-Based Load Balancing Methods.

1) M. Mishra and A. Jaiswal designed an ant-

based control system to address the issue of

LB in cloud contexts. To increase or decrease

various performance metrics, such as

network load, CPU usage, large memory, or

latency for clouds of varying sizes, we seek

to create an efficient algorithm for LB that

makes use of the simulated annealing

technique. It has been demonstrated that the

pheromone update is a useful tool for load

balancing.

2) Richa Chib, Er. V. Kaur and Dr. N. Dhillon

proposed a new method for evaluating an

optimized load balancer performance. By

altering the interval time, the provided

technique estimates the necessary measures

and is based on conditions that will give the

client high availability. They attempted to

prevent overloading in the suggested

technique and under VM loading.

3) Zheng-Tao Wu modified the basic ACO

algorithm to facilitate LB and job scheduling.

The ACO algorithm can efficiently find the

regional and international (global) ideal

solution as well as attain the quickest possible

implementation complexity time, according

to their results (S. Banerjee, 2009) By

modifying the pheromone formula, it will

predict the shortest time for task completion.

4) Shagufta Khan et al. applied the SALB

algorithm, studied current ACOs first, then

used ACO to create an efficient algorithm for

LB. Balancing the overall system load by

attempting to optimize or reduce the various

parameters is the primary contribution of the

work (A. Kaur,2020).

5) Soumya Banerjee et al. provided a

preliminary heuristic algorithm to implement

a modified ACO approach for the cloud

paradigm’s various scheduling and service

allocation mechanisms. The coefficient and

the ACO pheromone update mechanism are

changed. The likelihood of fulfilling the

inquiry has, moreover, modified scheduling

that has been used to converge, and this

modification helps to minimize the makespan

time (C. Udatha,2023).

Exploring Load Balancing: Issues, Methods and Strategic Outlook in Cloud Computing

657

Table 6: Examination of Ac0 Algorithm-Based Load Balancing Methods.

Title Author Method Key Points

ACO: A Survey of

load-balancing

M. Mishra and A.

Jaiswal

Heuristic algorithm

based on ACO

A pheromone update is a useful

and efficient tool for load

balancing.

It is approachable for routing in

the systematic network.

ACO-based LB

algorithm for cloud

computing

R. Chib, Er. V. Kaur

and Dr. N. Dhillon

ACO

algorithm to evaluate the

effectiveness of the

optimized load balancer

This technique avoids the

situations of overloading and

underloading of VMs.

Useful for both static and

dynamic Load balancing.

Application of ACO

in cloud computing

Zheng-T Wu

Modifying

ACO for task scheduling

It can find the best solution

worldwide.

It predicts the quickest time to

finish a task.

Effective scheduling

algorithm for LB

using

ACO in cloud

computing

S. Khan et al.

Scheduling Algorithm

(SALB)

Effective at locating the overfilled

nodes as quickly as possible.

To maintain mode balance while

maximizing resource utilization

and efficiency.

Cloud computing

initiative using a

modified ACO

framework.

S. Banerjee et al

The heuristic algorithm

uses modified ACO

Improved use of available

resources.

4.5 Load Balancing Using Particle

Optimization of Swarm (PSO)

The meta-heuristic optimization algorithm known as

PSO was motivated by the social behavior of fish and

birds. By mimicking the motions of a swarm of

particles within the search area, it is used to find the

best answers. Every particle serves as a possible

solution as well as modifies its location in response to

both its own as well as its neighbours’ experiences.

Table 7 shows the Comparison of The Enhanced PSO

Algorithm-Based Simulation Tool’s Performance.

1) Dr. A. Kaur, Dr. P. Singh, H. K. Toor, and B.

Singh provided the method of heuristic

optimization that is employed to improve the

decentralized load-balancing technique,

which distributes the load among each virtual

machine (VM). Additionally, the outcomes

are examined and contrasted with a

centralized load balancer in terms of

throughput and energy efficiency parameters

2) Chaitanya Udatha and Gondi Lakshneeswari

suggested an optimized multi-objective PSO

algorithm (LBIMOPSO) method for LB to

distribute tasks among the most appropriate

virtual machines (VMs) and manage load

consistently. It is a strong optimization

method that efficiently balances workloads in

a cloud computing environment while

considering several objective functions at

once.

3) R. M. Alguliyev et al. suggested the best way

to move workload-causing tasks from virtual

machines (VMs) that are overloaded to the

appropriate VMs. It aims to illustrate that

giving criteria weights results in a better

solution.

4) H. K. Nayak et al. suggested a hybrid strategy

for overcoming LB problems. In this method,

the author combined the Dragonfly and PSO

algorithms to get a better response time. It

combines investigating the dragonfly

algorithm and the strength of task scheduling

using particle swarm optimization, the

purpose of LB.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

658

Table 7: Comparison of the Enhanced PSO Algorithm-Based Simulation Tool’s Performance.

Title

Techniques

Tool Used

Response

Time

Improvement

PSO-based dynamic

load-balancing in

the cloud

environment

Decentralized LB using

enhanced PSO

CLoudSim 18% Improved

Adaptive load

balancing that

makes use of PSO

for cloud task

scheduling

LB improved

multiobjective PSO

(LBIMOPSO)

CloudSim 50.76% Improved

PSO-cloud

computing using a

cloud-based LB

approach

PSO-Time-

Based LB Alpha-

PSOTBLB

CloudSim,

Jswarm

0.211s

DPSO: A hybrid LB

strategy in cloud

computing that

combines the PSO

and dragonfly (DA)

algorithms

Hybrid approach

(DA+PSO)

CloudSim

50% over DA 66.67% over

PSO

5 EMERGING TRENDS IN

LOAD-BALANCING

The future of LB is being shaped by emerging

developments within the field of CC, which seeks to

optimize sustainability, scalability, efficiency, as well

as security. A closer look at some significant new

trends is provided here

5.1 Fog Computing

By incorporating the network edge into the ecosystem

of computing to enable decisions as near the data

sources as possible as well as feasible, the Fog model

is an expansion of the conventional cloud computing

paradigm. The use of such a computing model has

several advantages. Lag times between servers and

users, for instance, might be decreased (M.

Adhikari,et,al.2018) Table 8 shows the Fog

Computing Advantages And Difficulties.

Table 8: Fog Computing Advantages and Difficulties.

Advantages Difficulties

1. Reducing the

amount of bandwidth

used by the cloud.

1. Needs effective

coordination between

edge, fog, and cloud

nodes.

2. It enhances

performance for real-

time applications.

2. Risks to security

because of

decentralized

processing.

5.2 Blockchain Technology

Blockchain technology is becoming essential in

several industries to speed up and optimize

transactions by raising their degree of auditability,

dependability, and traceability. A distributed

immutable ledger that is set up in a decentralized

network and depends on cryptography to adhere to

security regulations makes up blockchain(M.

Exploring Load Balancing: Issues, Methods and Strategic Outlook in Cloud Computing

659

Adhikari,et,al.2018). Table 9 shows the Blockchain

Advantages And Difficulties.

Table 9: Blockchain Advantages And Difficulties.

Advantages Difficulties

1. It removes the

main causes of

failure.

1. Excessive computational

overhead.

2. Increases trust

and securit

2. Slower processing rates.

5.3 Machine and Deep Learning

Recently, there has been a lot of interest in AI because

of the enormous volume of data produced in recent

years, as well as the increase in processing power,

primarily from GPUs. Researchers and practitioners

of cloud computing can benefit from understanding

deep learning and machine learning algorithms and

prototypes (M. Adhikari and T. Amgoth,2018). Cloud

environments are increasingly using DL and AL for

intelligent load balancing. These methods make

decisions in real time, optimize resource allocation,

and forecast workloads using data-driven models.

• Predictive Analysis

• Reinforcement Learning

• Neural Networks

Table 10: AI Advantages and Difficulties.

Advantages Difficulties

1. Dynamically adjusts to

shifting workloads.

1. Training models

require large

datasets.

2. Enhances user experience

by making the best use of

available resources.

2. Computationally

costly.

New developments in load balancing emphasize

automation, decentralization, sustainability, and

intelligence. Cloud environments are becoming more

eco-friendly, scalable, and efficient thanks to

technologies like AI, Edge computing, Blockchain,

and Green cloud computing, which are

revolutionizing conventional load-balancing

techniques. (M. Adhikari, et, al.2018) Table 10 shows

the AI Advantages and Difficulties.

6 FUTURE RESEARCH

DIRECTION

In the first ten years of its existence, the idea behind

the CC has transformed the information technology

landscape, similar to the Internet, the Web, and the

actual computer. Large-scale diverse sensor networks

and IoT will produce enormous data streams for

archival purposes, management, and analysis, as well

as energy-cost-effective customized computer

services that must adjust to a range of hardware

gadgets while making adjustments for a number of

factors. The surveyed algorithms are found to

generally enhance energy protection, resource

utilization, as well as quality of service. Existing LB

algorithms have a number of drawbacks, including

static barriers, insufficient frequency control, resource

waste, and energy waste. As a result, there is much

room for improvement. Therefore, new approaches

that demand load-balancing according to carbon

emission, energy consumption, and support costs are

very promising. It is recommended that a number of

meta-heuristics be tested in real-world scenarios, such

as techniques that use ACO or PSO to demonstrate

their potential for use in the actual cloud. To get

around the shortcomings of the algorithms in use

today, the following work might be accomplished in

the future.

7 CONCLUSIONS

This survey has explored various cloud computing

elements that guarantee effective resource use, high

availability, and peak performance. Among the main

issues, load balancing is the primary problem since

burdening a gadget/device can have disastrous

consequences and render technology outdated.

Therefore, efficient resource utilization always

requires the use of an LB algorithm. The primary

objectives regarding LB are to fulfil user demands by

allocating the workload among several network nodes,

optimizing resource utilization, and increasing device

efficiency. Load balancing algorithms were explained

in this study, including dynamic load balancing, as

well as dynamic algorithms inspired by nature. More

efficient use of resources, a lower makespan, a higher

extent of the mismatch, and efficient migration of

tasks, as well as a shorter period, will all be made

possible eventually by the requirement to develop

completely self-governing dynamic LB algorithms.

The technology of CC itself has a long lifespan.

Among the key innovations, we can use it to carry out

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

660

crucial business functions. Over time, the

aforementioned innovations will completely improve

cloud computing. Effective and clever load-balancing

techniques will remain essential as cloud computing

grows to meet the growing needs of big data

processing, distributed applications, and cutting-edge

technologies like 5G and the Internet of Things(IoT).

The upcoming CC system generation will be greatly

influenced by ongoing research and innovation in this

region.

ACKNOWLEDGMENTS

I want to thank everyone who helped with this review

paper. I want to start by expressing my appreciation to

Mrs. Kavita Agrawal [Supervisor], whose knowledge,

insightful advice, and unwavering support helped to

shape the course of my study. I also thank Integral

University for its assistance in providing resources and

a favourable research environment. And at last, I’m

grateful to my friends and family for their unwavering

support as well as encouragement during the research

process. This work is an intellectual property of

Integral University videos the Manuscript

Communication no. IU/R&D/2025-MCN0003505.

REFERENCES

M. Shahid, N. Islam, M. Alam, M. Su'ud and S. Musa, "A

Comprehensive Study of Load Balancing Approaches

in the Cloud Computing Environment and a Novel

Fault Tolerance Approach," IEEE, vol. VIII, 2020.

D. Shafiq, N. Jhanjhi, A. Abdullah, and M. Alzain, "A load

balancing algorithm for the data centers to optimize

cloud computing applications," IEEE, p. 99, 2021.

F. Zabini, A. Bazzi, B. Masini and R. Verdone, "Optimal

performance versus fairness tradeoff for resource

allocation in wireless systems," IEEE Transactions on

Wireless Communications, vol. VI, no. 4, 2017.

J. Shah, K. Kotecha, S. Pandya, D. Choksi and N. Joshi,

"Load balancing in cloud computing: Methodological

survey on different types of algorithm," in International

Conference on Trends in Electronics and Informatics

(ICEI), 2017.

P. Kumar and R. Kumar, "Issues and challenges of load

balancing techniques in cloud computing: A survey,"

ACM Computing Surveys (CSUR), vol. LI, no. 6, pp. 1-

35, 2019.

A. Chaturvedi and A. Rashid, "Cloud Computing

Characteristics and Services: A Brief Review,"

International Journal of Computer Sciences and

Engineering, vol. II, pp. 421-426, 2019.

R. Khan and M. Ahmad, "Load balancing challenges in

cloud computing: a survey," in Proceedings of the

International Conference on Signal, Networks,

Computing, and Systems, 2016. K. A. Nuaimi, N.

Mohamed, M. A. Nuaimi and J. Al-Jaroodi, "A Survey

of Load Balancing in Cloud Computing: Challenges

and Algorithms," in Second symposium on network

cloud computing and applications, 2012.

M. A. Hossain and S. Roy, "Measuring the Performance on

Load Balancing Algorithms," Global Journal of

Computer Science and Technology, vol. XIX, no. 2,

2019.

A. Jain and R. Kumar, "A multi-stage load balancing

technique for cloud environment," in International

Conference on Information Communication and

Embedded Systems (ICICES), Chennai, 2016.

S. Afzal and G. Kavitha, "A Taxonomic Classification of

Load Balancing Metrics: A Systematic," in Indian

Engineering Congress, Udaipur, 2018.

R. Kaur and N. Ghumman, "Task-Based Load Balancing

Algorithm by Efficient Utilization of VMs in Cloud

Computing," in Big Data Analytics: Proceedings of CSI

2015, 2017.

N. Verma, B. N. Gohil, A. S, and K., "Load balancing in

Cloud Computing Environment using Modified

Genetic Algorithm," in 6th International Conference on

Information Systems and Computer Networks (ISCON),

2023.

S. Issawi, A. A. Halees and M. Radi, "An efficient adaptive

load balancing algorithm for cloud computing under

bursty workloads," Engineering, Technology & Applied

Science Research, vol. V, no. 3, 2015.

N. Pasha, A. Agarwal and R. Rastogi, "Round robin

approach for VM load balancing algorithm in cloud

computing environment," International Journal of

Advanced Research in Computer Science and Software

Engineering, vol. IV, no. 5, pp. 34-39, 2014.

S. A. Salman and M. K. Ahmed, "Load balancing

techniques in cloud computing: A review," Journal of

King Saud University –Computer and Information

Sciences, vol. VI, no. 1, pp. 223-250, 2021.

A. Aliyu and P. Souley, "Performance Analysis of a Hybrid

Approach to Enhance Load Balancing in a

Heterogeneous Cloud Environment," International

Journal of Advance in Scientific Research and

Engineering (IJASRE), vol. V, no. 7, 2019.

S. Liang, W. Jiang, F. Zhao, and F. Zhao, "Load Balancing

Algorithm of Controller Based on SDN Architecture

Under Machine Learning," Journal of Systems Science

and Information, vol. VIII, no. 7, pp. 578-588, 2020. J.

Kumar and A. K. Singh, "Workload prediction in cloud

using artificial neural network and adaptive differential

evolution," Future Generation Computer Systems, vol.

LXXXI, pp. 41-52, 2018.

A. Abbas, D. Suttr, C. Zoufal, A. Lucchi, A. Figalli and S.

Woerner, "The power of quantum neural networks,"

Nature Computational Science, vol. I, pp. 403-409,

2021.

A. K. Singh, D. Saxena, J. Kumar and V. Gupta, "A

Quantum Approach Towards the Adaptive Prediction

of Cloud Workloads," IEEE Transactions on Parallel

Exploring Load Balancing: Issues, Methods and Strategic Outlook in Cloud Computing

661

and Distributed Systems (, vol. XXXII, no. 12, pp.

2893-2905, 2021.

R. Mishra and A. Jaiswal, "Ant colony Optimization: A

Solution of Load balancing in Cloud," International

journal of Web & Semantic Technology, vol. III, no. 2,

p. 33, 2012.

R. Chib, V. Kaur and N. Dhillon, "Load Balancing

Algorithm For Cloud Computing Using Ant Colony

Optimization," Journal of Emerging Technologies and

Innovative Research, vol. V, no. 10, 2018.

Z. T. Wu, "Application of Ant Colony Optimization in

Cloud Computing Load Balancing," in International

Conference on Intelligence Science, 2017.

S. Khan and N. Sharma, "Effective Scheduling Algorithm

for Load balancing (SALB) using Ant Colony

Optimization in Cloud Computing," International

Journal of Advanced Research in Computer Science

and Software Engineering, vol. IV, no. 2, pp. 966-973,

2014.

S. Banerjee, I. Mukherjee and P. K. Mahanti, "Cloud

computing initiative using modified ACO framework,"

World Academy of Science Engineering And

Technology, vol. III, 2009.

A. Kaur, P. Singh and H. K. Toor, "Particle Swarm

Optimization (PSO) based Dynamic Load Balancing in

Cloud," International Journal of Computer Science

Engineering (IJCSE), vol. IX, no. 2, pp. 130-136, 2020.

C. Udatha and G. Lakshmeeswari, "An Adaptive Load

Balancing using Particle Swarm Optimization for

Cloud Task Scheduling," International Journal of

Engineering Trends and Technology, vol. LXXI, no. 9,

pp. 36-45, 2023.

R. Alguliyev, Y. Imamverrdiyev and F. Abdullayeva,

"PSO-based load balancing method in cloud

computing," Automatic Control and Computer

Sciences, vol. LIII, pp. 45-55, 2019.

S. Mohapatra, S. Mohanty, H. Nayak, M. Mallick, J.

Ramesh, and K. Dudekula, "DPSO: A Hybrid

Approach for Load Balancing using Dragonfly and

PSO Algorithm in Cloud Computing Environment,"

EAI Endorsed Transactions on Internet of Things,

2024.

R. Buyya et al.., "A Manifesto for Future Generation Cloud

Computing: Research Directions for the Next Decade,"

ACM Computing Surveys, vol. LI, no. 5, 2018.

M. Adhikari and T. Amgoth, "Heuristic-based load-

balancing algorithm for IaaS cloud -," Future

Generation Computer Systems, vol. LXXXI, pp. 156-

165, 2018.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

662