Комплексний огляд технік балансування навантаження для віртуальних машин
Cloud computing is an emerging technique with remarkable features such as scalability, high flexibility, and reliability. Since this field is growing exponentially, more users are attracted to fast and better service. Virtual Machine (VM) allocation plays a crucial role in cloud computing optimizati...
Saved in:
| Date: | 2023 |
|---|---|
| Main Authors: | , |
| Format: | Article |
| Language: | English |
| Published: |
The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute"
2023
|
| Subjects: | |
| Online Access: | https://journal.iasa.kpi.ua/article/view/267280 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Journal Title: | System research and information technologies |
| Download file: |  |
Institution
System research and information technologies| _version_ | 1866302876189458432 |
|---|---|
| author | Sansanwal, Suman Jain, Nitin |
| author_facet | Sansanwal, Suman Jain, Nitin |
| author_sort | Sansanwal, Suman |
| baseUrl_str | http://journal.iasa.kpi.ua/oai |
| collection | OJS |
| datestamp_date | 2024-02-01T21:03:07Z |
| description | Cloud computing is an emerging technique with remarkable features such as scalability, high flexibility, and reliability. Since this field is growing exponentially, more users are attracted to fast and better service. Virtual Machine (VM) allocation plays a crucial role in cloud computing optimization; hence, resource distribution is not impacted by machine failure and is migrated with no downtime. Therefore, effective management of virtual machines is necessary for increasing profit, energy-saving, etc. However, it could utilize the virtual machine resources more efficiently because of the increased load, so load balancing is more concentrated. The predominant purpose of load balancing is to balance the available load equally among the nodes to avoid overloading or underloading problems. The present study conducted an extensive survey on virtual machine placement to describe the application of prediction algorithms and to provide more efficient, reliable, high response, and low overhead VM placement. Furthermore, the survey attempted to overview the challenges in load balancing in VM placement and various ideas of state-of-the-art techniques to resolve the issues. |
| doi_str_mv | 10.20535/SRIT.2308-8893.2023.4.10 |
| first_indexed | 2025-07-17T10:28:00Z |
| format | Article |
| fulltext |
Suman Sansanwal, Nitin Jain, 2023
Системні дослідження та інформаційні технології, 2023, № 4 135
UDC 62-50
DOI: 10.20535/SRIT.2308-8893.2023.4.10
A COMPREHENSIVE SURVEY ON LOAD BALANCING
TECHNIQUES FOR VIRTUAL MACHINES
SUMAN SANSANWAL, NITIN JAIN
Abstract. Cloud computing is an emerging technique with remarkable features such
as scalability, high flexibility, and reliability. Since this field is growing exponen-
tially, more users are attracted to fast and better service. Virtual Machine (VM) allo-
cation plays a crucial role in cloud computing optimization; hence, resource distribu-
tion is not impacted by machine failure and is migrated with no downtime.
Therefore, effective management of virtual machines is necessary for increasing
profit, energy-saving, etc. However, it could utilize the virtual machine resources
more efficiently because of the increased load, so load balancing is more concen-
trated. The predominant purpose of load balancing is to balance the available load
equally among the nodes to avoid overloading or underloading problems. The pre-
sent study conducted an extensive survey on virtual machine placement to describe
the application of prediction algorithms and to provide more efficient, reliable, high
response, and low overhead VM placement. Furthermore, the survey attempted to
overview the challenges in load balancing in VM placement and various ideas of
state-of-the-art techniques to resolve the issues..
Keywords: virtual machine allocation, load balancing, cloud computing, overload-
ing, physical machine, data center.
INTRODUCTION
Cloud computing is now becoming vital for hosting several IT services that
provide various on-demand VR (Virtual Resources). The cloud service providers
used large-scale DC (data center) with more physical machines. Virtualization is
more beneficial in data centers for providing the VM comprising a software layer
known as a VVM Monitor. This VMM enables the controlling of shared physical
machine resources, thereby increasing VM security but accommodating multiple
VM in a single physical machine remains a challenging problem. Due to this
problem, there is a chance of overutilizing PM, degrading it, or wasting high-cost
resources.
Further, the power consumption of cloud DC mainly occurs by physical ma-
chines, so proper VM placement associated with dynamic management greatly
mitigates DC power consumption, improving the profit and throughput and pre-
venting SLA violations. However, VM placement employs a widespread and ex-
pensive VM migration process. If improper placement occurs, it will lead to the
destruction of data centre performance. Furthermore, balancing the request’s
workload and allocating appropriate tasks to the appropriate VM is also consid-
ered challenging. Hence load balancing is a crucial factor to be considered with
the increasing requests and impulsive arrival patterns. Load balancing is the even
classification of the task processed in-between more CPUs, storage devices, and
network links which deliver fast service with more efficiency. This is obtained
Suman Sansanwal, Nitin Jain
ISSN 1681–6048 System Research & Information Technologies, 2023, № 4 136
using hardware/software devices and multiple servers that appear as a computer
clustering. In addition, load balancing improves the efficiency of distributed or
parallel systems through load redistribution. Load balancing algorithms are classi-
fied into static and dynamic algorithms. The static algorithm is simple and needs
minimized runtime overhead, whereas a dynamic system is utilized in most of the
modern load-balancing approaches due to its flexibility and robustness. Likewise,
there exist four kinds of load-balancing policies, which are location policies,
transfer policies, selection policies, and information policies. Other such objectives
of load balancing include reducing carbon emission and energy consumption,
resource provisioning, avoiding bottlenecks, and achieving QoS requirements.
To overcome the prevailing limitations and obtain end-user satisfaction,
high-quality and effective methodologies must be adopted to support the optimi-
zation of VM load balancing. Therefore, the study’s main contribution is to pro-
vide a comprehensive survey of the existing methods dealing with virtual machine
load balancing by the factors affecting the cloud computing process.
Objective:
To analyze several virtual machine placement and load balancing
techniques in the existing literature.
To overview the prevailing challenges in load balancing in virtual
machine processing and to provide a comprehensive outlook to rectify the issues.
To outlook the recent trends for the optimization of load balancing in
virtual machine allocation.
The paper is organized as follows: Section 2 deals with the predictive virtual
machine placement methods with various algorithms, and Section 3 reviews pre-
vailing load balancing techniques such as static and dynamic methods. Section 4
summarizes the advantages of load balancing in virtual machine allocation, and
Section 5 overviews the performance metrics for evaluating load balancing in vir-
tual machines. Section 6 provides the recent trends in this concept, and Section 7
deliberates on the challenges and research gaps of load balancing for virtual
machines. Followed by Section 8 concludes the work.
PREDICTIVE VIRTUAL MACHINE PLACEMENT METHODS
Various predictive virtual machine placement methods are designed and sug-
gested for the CC environment. Besides, implementing all these methods for en-
hancing the placement process of virtual machines by utilizing historical data.
The predictive methods for the VM placement are classified as the following [1].
Ensemble-based scheme.
Hybrid scheme.
Exponential smoothing predictor-based scheme.
Dynamic programming-based scheme.
Grey model-based scheme.
Fractal based schemes.
Bayesian-based scheme.
Neural network-based scheme.
SVM based scheme.
Queuing based scheme.
Markov based scheme.
A comprehensive survey on load balancing techniques for virtual machines
Системні дослідження та інформаційні технології, 2023, № 4 137
Hidden Markov Model based scheme.
ARIMA model-based scheme.
Regression-based scheme.
The following are a few existing virtual placement schemes that utilize the
above-said algorithms. This study [2] attempted to predict resource requirements
for virtual machines to improve the process of virtual machine placement. Be-
sides, the study indicated that the time-reliable hidden Markov model replicated
the properties of CPU utilization data and determined the CPU’s future usage.
Nevertheless, the study applied only univariate normal distribution to determine
resources, whereas the multivariate normal distribution to determine multiple re-
sources could be useful. Likewise, [3] provided a Markov predicting framework
for forecasting the future under-utilized/over-utilized physical machines and pre-
venting unnecessary and immediate migration of virtual machines. Besides, this
study utilized the cloud sim toolkit evaluated using random forest and Planet Lab
datasets. Finally, the current usage of the CPU of every physical machine has
been compared with upper and lower thresholds for recognizing the status.
Moreover, determine the future state of physical machines by implementing
the Markov model. This study [4] introduced a framework for identifying the rela-
tionship of resources amid virtual machines by utilizing ARIMA-based determi-
nations. Further, the study analyzed resource utilization after the placements of
two virtual machines on the same physical machine, and also the study named this
system an affinity model. Similarly, this study [5] implemented automata for en-
hancing the usage of resources as well as mitigating the usage of power. Further-
more, this study considered the variations in user demands for estimating over-
loaded physical machines. Due to the prevention of physical machine overload,
this system improves the utilization of resources, mitigates the amount of migra-
tion, and shuts the idle physical machines to mitigate the utilization of power. Fi-
nally, the study stated that this method was executed in the cloudsim toolkit by
utilizing Planet Lab dataset. Nevertheless, this cannot detect underutilized physi-
cal machines.
REVIEW OF PREVAILING LOAD BALANCING METHODS
This study surveyed the literature on prevailing load-balancing methods and com-
prehensively reviewed certain studies. Besides, the load balancing methods are
segregated into dynamic and static, based on the system’s state. Load balancing is
needed to improve resource utilization, reducing the completion and response
time for the tasks on the cloud. This study [6] suggested a method in which it con-
sidered QoS, number of migrations as well as response time as the parameters of
load balancing. Further, tasks with less priority have been transferred from one
virtual machine to another when overloaded with virtual machines. This method
can be improvised with other algorithms like ACO and PSO.
Static load balancing methods. The static load balancing method doesn’t
require knowledge of a system’s current state; it requires knowledge of the system
resources like processing power, storage capacity, memory, and execution time in
advance. Besides, the static load balancing methods don’t allow resource alloca-
tion at execution time. Also, these methods are easy to execute and implement,
but they are beneficial to small networks or systems with a minimum amount of
resources. On the other hand, as they don’t consider the present state of the sys-
tem, these methods aren’t beneficial for computing systems that perform distrib-
uted computing. Moreover, they need to permit the detection of connected server
machines at the execution time, thereby leading to uneven resource distribution.
Suman Sansanwal, Nitin Jain
ISSN 1681–6048 System Research & Information Technologies, 2023, № 4 138
Dynamic load balancing methods. Since the static load balancing tech-
niques are not suitable for the distributed computing system, the dynamic load
balancing methods are suitable in a cloud computing environment. The following
are different load balancing methods, which rely on the criteria of the load balancer:
Cluster-based load balancing.
Task-based load balancing.
Agent-based load balancing.
Hybrid load balancing.
Natural phenomena based on load balancing.
General load balancing.
Cluster-based load balancing.
This study [7] addressed a heuristic method for load balancing based on
(LB-BC) Bayes and Clustering for overcoming the difficulties of prevailing load
balancing techniques. This technique is based on Bayes’ theory and has accom-
plished long-term load balancing. This computes the posterior probability of the
physical hosts and integrates with clustering for picking an optimal host. Further,
this considered the parameters like load balancing effect, standard deviation, and
the number of requested tasks. Then, it was compared with the dynamic load bal-
ancing, leading to increased time and minimal standard deviation. This method
only works in localized areas, but further enhancement can be made for working
in a real-time environment and a wide area network. This study [8] presented a
cluster-based method for improvising intercloud communication in real-time and
dynamic multi-media for load balancing. This method has a two-step process. The
first step is to develop the cluster to monitor the activities, handle platform diffi-
culties, and meet the satisfactory quality of service and demands for hosts based
on a hello-packet broadcast for all the servers. In the second step, it decides on
transfer job requests. When this method was compared with HFA, WCAP, and
ant colonies, the suggested method produced an improved response time. In addi-
tion, this method could be improvised for a real-time environment, in which the
intermediate nodes are congested, and owing to reduce the data loss because of
congestion by utilizing communication jobs instead of computation jobs.
This study [9] presented a cluster-based load-balancing method for overcom-
ing load distribution issues. Besides, this integrated the concept of KUHN and
genetic algorithm and created a task allocation strategy by grouping the tasks into
clusters and distributing them in a cooperating node. As a result, this method pro-
vided improvised task distribution and response time among data center nodes.
Similarly, this study [10] created a hierarchical model to self-schedule the
schemes for improving the scalability and load balancing of the cloud system.
Besides, this method can extract in a heterogeneous and homogeneous environ-
ment. In addition, this study has implemented the schemes on a large scale by
utilizing various computation applications. Finally, the outcomes of the study de-
picted improvised scalability and overall performance, as well as decreased com-
munication overhead. The further analysis deals with a testing algorithm for
large-scale loops and clusters with dependencies.
Task-based load balancing. This study [11] presented a network-aware task
placement method for reducing task completion time, data cost, and transmission
time. The study stated that the three challenges faced by tasks are the availability
of resources dynamically changes resulting in access over time; data fetching time
relies on the task’s location and size; the load on the path significantly impacts the
data access latency. Therefore, the study must consider loading over the path dur-
A comprehensive survey on load balancing techniques for virtual machines
Системні дослідження та інформаційні технології, 2023, № 4 139
ing scheduling to minimize this latency. The study’s outcomes depicted that the
suggested method has significantly reduced the task’s completion time and in-
creased resource utilization.
This study [12] suggested a scheduling technique for reducing the resource
competition between high device load and tasks based on the weighted random
scheduling method. The tasks are assigned by considering parameters such as
communication delay, time, and cost. Besides, the study was analyzed with
MATLAB software by utilizing workflow for generating the dataset. Also, the
study analyzed the dataset, which included a large set of tasks with transmission
delay, cost, and time. Moreover, the study considered device dependency, task
arrival time, and task structure. The study’s outcomes depicted that multiple
schedules have seen improvement in parameters like execution cost and task
completion for the devices. Nevertheless, it still needs to calculate the optimal
value for parameters that could be improvised in further analysis.
Agent-based load balancing. The multi-agent-based load-balancing frame-
work helps increase resource utilization [13]. This executed both the receiver
originate method, as well as the sender, originated method for reducing the wait-
ing time of tasks and also for assuring SLA. This method incorporated the agents
like NA (Negotiator Ant) agent, DCM (Datacenter Monitor) Agent, as well as
VMM (Virtual Machine Monitor) Agent. Among these, the virtual machine moni-
tor agent supports every virtual machine in the system and retains the information
on bandwidth, CPU, and memory by utilizing virtual machines for monitoring the
load. Besides, the datacentre monitor agent executes information policy by utiliz-
ing the available information from the virtual machine monitoring agent and cate-
gorizing the virtual machines relying upon various characteristics. Also, this initi-
ates the negotiator and agent that moves to various other data centers for
identifying the available virtual machines’ status. From the experimental analysis,
the study stated that the suggested method was more effective, improving the re-
sponse time and reducing the makespan time.
The (SVLL) selection of virtual machines with the least load balancing tech-
nique for the distribution of tasks increased the cloud computing performance
[14]. This model computes a load of every virtual machine and assigns tasks to
evaluate based on the virtual machine’s load rather than the number of tasks as-
signed to virtual machines. Besides, the study implemented the SVLL method
with various task scheduling methods like shortest job first and first came first
serve methods, in which the outcomes of the study denoted that the suggested
method has improvised in total finishing time and total waiting time. In addition,
this method was employed with basic task scheduling methods for better results.
This study [15] developed a load-balancing method by integrating round-
robin features and shortest-job-first scheduling algorithms. This method stores
long and short tasks in separate queues and utilizes dynamic task scheduling
quantum to balance waiting time among the tasks. Besides, this study has taken
into account the issues of starvation as well as throughput. Also, they executed the
experiment on the cloud tool. As a result, the experimental analysis showed that
response time, waiting time, and the turnaround time was reduced. In addition to
that, long-task starvation was also minimized. Nevertheless, the task quantum was
not efficient in balancing the tasks, but it could be improvised in further analysis.
The hybrid load-balancing method. This study [16] employed a hybrid al-
gorithm for optimizing the system’s performance by integrating throttled and
round-robin load balancing methods with a service-proximity broker and per-
formance-optimized service broker algorithm. Besides, the study suggested one
Suman Sansanwal, Nitin Jain
ISSN 1681–6048 System Research & Information Technologies, 2023, № 4 140
load balancing and three service broker methods. The study denoted them as CA
(Cost Aware) and LA (Load Aware) algorithms for high utilization of resources.
However, although the LA algorithm offers low processing time, it can generate
high costs, whereas CA reduces cost. Moreover, the service broker algorithm decides
on the server to users’ requirements, which might increase cost or processing time.
In contrast, the service proximity algorithm decides on the data center near
to client’s region. Finally, the study integrated all the algorithms, and the out-
comes of the study denoted that response time and processing time have signifi-
cantly reduced. Nevertheless, further analysis deals with the improvisation of sys-
tem performance. The development of an efficient CLB (Cloud Load Balancing)
framework is needed to overcome the server failure response in the event of sev-
eral user requests. Several studies have developed a framework that considers the
loading and server processing for minimizing the server problems for handling
various computation requests. Also, they presented a load-balancing method for
virtual and physical web servers to preserve the information regarding computing
power, priority, and server loading. Even though this framework provides high
scalable performance, it can increase response time.
COMPARATIVE ANALYSIS OF STATIC AND DYNAMIC LOAD BALANCING
TECHNIQUES
Table provides a comprehensive comparative analysis of the existing load-
balancing algorithms.
Comparative Analysis between The Existing Load Balancing Algorithms
S.
No
Type of
Load
Balancing
algorithm
in VM
Load Balancing
Algorithm
Parameters
enhanced Merits Demerits
1
Weighted-round
robin
algorithm [17]
Waiting and
response time
Utilize all resources in
a balanced manner.
Ensuring fairness in
every allocation
Execution time
Prediction is not
possible. High
Migration time
2
Opportunistic load-
balancing algorithm
[18]
User
discomfort cost
and reduction
The end-user achieves
better accuracy and
comfort maximization
Comfort maximi-
zation might lead
to raised costs and
energy
3
Static
Software-Defined
Networking based
load-balancing al-
gorithm [19]
Cost, response
time, and
scalability
Effective user
request processing
Increased
energy
consumption
4 Ant colony
algorithm [20]
Makespan,
response time,
scalability
Good scalability, Fault
tolerance, and obtain-
ing load
balancing for
Complex networks.
High power
consumption. Less
throughput
5
Deadline-
constrained based
dynamic
load-balancing al-
gorithm [21]
Task
rejection ratio,
makespan
Increases the
utilization ratio
Increased
consumption
of cost
6
Dynamic
Honey-bee
foraging
algorithm [22]
Response time,
throughput
Less waiting time
and Increased
system diversity
High
response time
Less throughput
A comprehensive survey on load balancing techniques for virtual machines
Системні дослідження та інформаційні технології, 2023, № 4 141
BENEFITS OF LOAD BALANCING IN VIRTUAL MACHINES
Ideally, these solutions can be implemented when performing the placement of
virtual machines. Decreasing the number of physical machines as well as consoli-
dating virtual machines could be utilized for solving cloud-spot issues. Reducing
the migrations of virtual machines by predicting future workloads will prevent
unnecessary migrations of virtual machines. Future pages could be identified by
mitigating transmitted pages by properly predicting the workload of applications.
Consequently, the number of transmitted pages could be diminished in the pre-
copy approach.
The load balancer offers flexibility for balancing the server’s workload by
traffic distribution across multiple servers. Further, load-balancing targets mimic
a software infrastructure via Virtualization. This runs physical load-balancing
software on VM. In addition, availability, performance, scalability, and reliability
are the major metrics of load balancing.
Availability. The mechanism of load balancing assures an efficient offer of
service. Moreover, the loads will be effectively distributed in terms of server un-
availability.
Performance. An effective load balancing provides cloud applications as
well as cloud services for responding faster when compared to the average com-
pletion time. In addition, execution time is also decreased via effective compres-
sion methods and catching mechanisms.
Scalability. The major benefit of the load-balancing technique is that some
servers can be easily included without any disturbance, and the applications can
be smoothly performed via the load-balancing servers.
Reliability. The reliability of cloud services was secured by the redundancy
of servers in which the applications could be hosted. Even in failure cases, the
cloud-serving resources will function, and its services will be redirected to other
locations in the cloud.
PERFORMANCE METRICS IN THE EVALUATION OF LOAD BALANCING
FOR VIRTUAL MACHINES
Various virtual machine load balancing metrics are present for assessing load
balance performance. These metrics were reflected in diverse task scheduling be-
havior. The following are the load balancing metrics.
Load variance. Consider that there exists n number of hosts in the data cen-
ter. The usage of host i can be expressed as )( ihostU , whereas the average usage
of every host can be calculated as
i
n
i
t hostUavg
12
1
)( .
Makespan time. Makespan time is known as the longest-processing time on
every host. Also, it is a normal criterion for accessing scheduling algorithms. Re-
taining load balance is for shortening the makespan time.
Overloaded hosts. The overload threshold can be denoted as )( tUT , for n
number of hosts, the host utilization can be expressed as )( ihostU , and the over-
load hosts is expressed as the following, )())(( it hostUUTNum .
Suman Sansanwal, Nitin Jain
ISSN 1681–6048 System Research & Information Technologies, 2023, № 4 142
Throughput. Throughput deals with system performance. A maximum
number of tasks are executed to accomplish high performance within the minimal
completion time.
SLA violations. Similarly, this also deals with the performance of the sys-
tem. The virtual machines can’t fetch adequate resources from the host, so the
host isn’t well-balanced. Thus, SLA violations must be reduced.
Turnaround time. Turnaround time is defined as the time systems take
from the request submission to a response from the server. And turnaround time
can be calculated as
Turnaround time Tt CC .
From the above equation Ct refers to completion time, and GT refers to gen-
eration time.
Overhead. Generally, overhead occurs because it increases the communica-
tion cost or takes more time to migrate from one virtual machine to another. Good
load-balancing algorithms will decrease the overhead.
Resource usage. Good performance usually deals with the proper resource
usage among nodes. This will be beneficial for measuring if the nodes are under-
loaded or overloaded.
Fault tolerance. This enhances the systems such that the single failure point
doesn’t impact the entire system. Besides, the load balancing algorithm must be
designed in a way where the failure of one node must not affect the system.
Response time. Generally, response time is the time taken by load balancing
techniques to users. Lesser response time indicates better system performance.
Therefore, load balancing will be more beneficial for the entire cloud by decreas-
ing the response time of cloud servers and task scheduling issues; the following
articles discuss the response time in virtual machines.
This study [23] suggested TMA (Throttled Modified Algorithm) improves
the response time of virtual machines on CC (Cloud Computing) to improvise a
performance. Besides, this study simulated the suggested method with the clouds
tool; the evaluated outcomes showed improved processing time and response
time.
In this study [24], a firefly load balancing technique was utilized to solve the
load imbalance problems in a cloud server to enhance the learners’ user experi-
ence. The suggested method needs a cloud-server mapping method for various
virtual machine methods, ensuring the users receive the content without delay.
From the experimental analysis, the study stated that, compared to the existing
method, the suggested method showed less response time.
RECENT TRENDS OF LOAD BALANCING IN VIRTUAL MACHINE
ALLOCATION
This study [25] suggested that response time was similar to execution time in eve-
ry task, and this parameter should be minimized. This determines the virtual ma-
chine status based on the current load. Later, the tasks are eliminated from the
machine with additive load, which depends on the virtual machine’s condition.
Finally, it will be transferred to the appropriate VM, which is the criteria to assign
A comprehensive survey on load balancing techniques for virtual machines
Системні дослідження та інформаційні технології, 2023, № 4 143
tasks to virtual machines based on the least distance. The outcomes of the cloud-
sim tool evaluation showed that response time was improved compared to exist-
ing algorithms. Additionally, the degree of load imbalance has also seen some
improvements.
The main aim of task scheduling incorporates scheduling resources and re-
ducing the schedule’s objective. This study [26] suggested a mean grey-wolf op-
timization technique to enhance the system’s performance and reduce scheduling
problems. The primary objective of this study is to reduce energy consumption
and makespan time. This was evaluated by utilizing the cloudsim tool. The study
showed that the suggested algorithm had better results than the prevailing
methods.
This suggested method in this study [27] attempted to avoid SLA violations
via power optimization and optimal cloudlet by reducing the migrations of virtual
machines. Besides, the SLA reduction system incorporated three parts a schedul-
ing algorithm, a MinVM scheduling algorithm, and a credit-based virtual machine
migration algorithm. When considering the scheduling algorithm, it efficiently
schedules the cloudlets to VMs based on the host’s processing time. Likewise, the
MinVM scheduling algorithm schedules the cloudlets to VMs based on counts of
cloudlet allocation to every virtual machine. And the credit-based algorithm util-
izes the virtual machine’s credit to take virtual machine migration.
CHALLENGES AND RESEARCH GAP
The most challenging task in virtual technology is virtual machine placement on
the physical machine under optimal conditions in cloud-data centers. Further, the
virtual machine placement can result in managing resources and preventing the
wastage of resources. Minimizing energy consumption, cost reduction, utilization
of resources, and presentation of best QoS are significant challenges in the cloud
computing environment. Since only a few studies focus on privacy and security
issues, in further analysis, security is a crucial factor that must be focussed on.
Besides, attackers can steal the secrets from other tenants by utilizing side-
channel attacks based on shared resources since the virtual machines from various
tenants might be located at one physical machine, thereby threatening data secu-
rity in a cloud computing environment. The following are certain limitations that
should be considered,
The forecasting approaches employed in predictive virtual machine place-
ment schemes could be enhanced to better deal with non-linear and linear loads.
Moreover, the predictive virtual machine placements in the multi-cloud
and multi-site cloud environments must be studied for further analysis.
Even though dynamic power management can be implemented to
improvising DCs energy efficiency, only a few studies have suggested this
approach in their literature.
One of the significant problems avoided by various studies is DDoS
attacks that could be originated from malicious virtual machines by uplifting the
resource demands and introducing several unnecessary virtual machine migrations.
The integration of predictive virtual machine placement methods with
prevention and intrusion detection systems must be investigated to recognize the
true demands and increase the DC’s security.
Suman Sansanwal, Nitin Jain
ISSN 1681–6048 System Research & Information Technologies, 2023, № 4 144
Studies must design and apply low-overhead placement methods in
developing technologies like mobile and cloudlets in the future.
In future studies, context-aware virtual machine placement must be
designed for the environment, like predicting mobile patterns, vehicular CC, and
connectivity problems.
In recent years, cloud computing has seen rapid growth and advanced re-
search in computation and data based on practical and theoretical aspects. Never-
theless, cloud computing researchers face several problems in which load balanc-
ing is more challenging and needs special attention. Besides, issues like user QoS
(Quality of Service) satisfaction, virtual machine security, resource usage, and
virtual machine migration must be considered to find a feasible solution to im-
prove resource utilization. Additionally, various problems like the migration of
virtual machines, resource utilization, QoS satisfaction, and migration of virtual
machines need equal attention for finding the optimal solution to enhance the op-
timal solution to improve the utilization of resources.
The following are certain load-balancing problems.
Geographically distributed nodes. Generally, the data centers in the cloud
are geographically-distributed. In these data centers, for effective system execu-
tion according to the request of users, the spatially distributed nodes were treated
as a single location system. Besides, certain load balancing methods were de-
signed for a small area. For example, they don’t consider communication delay,
network delay, and the distance between distributed resources, users, and comput-
ing nodes. Nevertheless, the nodes situated at various locations are challenging
since these algorithms are unsuitable for these environments. Therefore, load-
balancing techniques for distantly located nodes must be considered [28].
Migration of Virtual Machines. Virtualization allows for the creation of
numerous virtual machines on one physical machine. As a result, virtual machines
are generally independent and possess various configurations. Besides, if the
physical machine is overloaded, certain virtual machines must migrate to a distant
location using the virtual machine migration load balancing method [29].
Heterogeneous Nodes. During earlier research in load balancing, several
studies have theorized about homogeneous nodes. But, usually, in cloud comput-
ing, users’ requirements dynamically change, which needs executing time for ef-
ficient resource utilization and decreasing response time. Thus, introducing an
effective load-balancing method for the heterogeneous environment is more chal-
lenging [30].
Storage management. Cloud storage solved the issues of the conventional
storage system, which required higher hardware costs and personnel management.
Further, the cloud allows users to store data heterogeneously without access is-
sues as there is a rapid increase in cloud storage, data replication for data consis-
tency, and effective access. However, because of duplicate storage policies, full
data replication is ineffective. The partial replication could be adequate. However,
there are certain issues in the dataset’s availability, and there might be increased
complexities in load balancing methods [31].
Scalability of the load balancer. The scalability and on-demand availability
of cloud services allow users to access the services for rapidly scaling up or scaling
down. Therefore, a load balancer must consider rapid variations by system topol-
ogy, storage, and computing power to efficiently facilitate these variations [32].
Complexity of algorithms. In a cloud computing environment, usually, the
algorithms must be easier and simple to implement. Besides, complex algorithms
may diminish the efficiency and performance of cloud systems [33].
A comprehensive survey on load balancing techniques for virtual machines
Системні дослідження та інформаційні технології, 2023, № 4 145
CONCLUSION
In general, a cloud indicates a distinct IT environment designed for the proper
functioning of remotely providing scalable and measurable IT resources. The paper’s
main objective is to consolidate the prevailing VM placement and load-balancing
methodologies. Further various challenges to the enhancement of effective VM
and load-balancing algorithms are also discussed. This survey lets the users look
at recent trends in VM placement and load balancing, enabling them to frame an
effective research methodology with maximum profit and minimum cost.
Declaration. I confirm that this work is original and has not been published
elsewhere, nor is it currently under consideration for publication elsewhere.
Acknowledgments
None.
Funding
Any organization/institute/agency did not fund this research work.
Competing Interests
None of the authors have any competing interests in the manuscript.
Availability of data and material
Not Available.
Code availability
Not Available.
Compliance with ethical standards
Ethical statement
No human participants or animals are involved in this research.
Consent statement. I confirm that any participants (or their guardians if un-
able to give informed consent, or next of kin, if deceased) who may be identifi-
able through the manuscript (such as a case report) have been allowed to review
the final manuscript and have provided written consent to publish.
REFERENCE
1. M. Masdari and M. Zangakani, “Green cloud computing using proactive virtual ma-
chine placement: challenges and issues,” Journal of Grid Computing, pp. 1–33, 2019.
2. H.L. Hammer, A. Yazidi, and K. Begnum, “An inhomogeneous hidden Markov
model for efficient virtual machine placement in cloud computing environments,”
Journal of Forecasting, vol. 36, pp. 407–420, 2017.
3. S.B. Melhem, A. Agarwal, N. Goel, and M. Zaman, “Markov prediction model for
host load detection and VM placement in live migration,” IEEE Access, vol. 6,
pp. 7190–7205, 2017.
4. X. Fu and C. Zhou, “Predicted affinity based virtual machine placement in cloud comput-
ing environments,” IEEE Transactions on Cloud Computing, vol. 8, pp. 246–255, 2017.
5. M. Ranjbari and J.A. Torkestani, “A learning automata-based algorithm for energy
and SLA efficient consolidation of virtual machines in cloud data centers,” Journal
of Parallel and Distributed Computing, vol. 113, pp. 55–62, 2018.
6. K.R. Babu and P. Samuel, “Enhanced bee colony algorithm for efficient load balanc-
ing and scheduling in cloud,” in Innovations in bio-inspired computing and applica-
tions. Springer, 2016, pp. 67–78.
7. J. Zhao, K. Yang, X. Wei, Y. Ding, L. Hu, and G. Xu, “A heuristic clustering-based
task deployment approach for load balancing using Bayes theorem in cloud envi-
ronment,” IEEE Transactions on Parallel and Distributed Systems, vol. 27,
pp. 305–316, 2015.
Suman Sansanwal, Nitin Jain
ISSN 1681–6048 System Research & Information Technologies, 2023, № 4 146
8. B. Kang and H. Choo, “A cluster-based decentralized job dispatching for the large-
scale cloud,” EURASIP Journal on Wireless Communications and Networking, vol.
2016, pp. 1–8, 2016.
9. F. Zegrari, A. Idrissi, and H. Rehioui, “Resource allocation with efficient load bal-
ancing in cloud environment,” in Proceedings of the International Conference on
Big Data and Advanced Wireless Technologies, 2016, pp. 1–7.
10. Y. Han and A. T. Chronopoulos, “Scalable loop self-scheduling schemes for large-
scale clusters and cloud systems,” International Journal of Parallel Programming,
vol. 45, pp. 595–611, 2017.
11. H. Shen, A. Sarker, L. Yu, and F. Deng, “Probabilistic network-aware task place-
ment for mapreduce scheduling,” in 2016 IEEE International Conference on Cluster
Computing (CLUSTER), pp. 241–250.
12. Y. Xin, Z.-Q. Xie, and J. Yang, “A load balance oriented cost efficient scheduling
method for parallel tasks,” Journal of Network and Computer Applications, vol. 81,
pp. 37–46, 2017.
13. S. Keshvadi and B. Faghih, “A multi-agent based load balancing system in IaaS cloud
environment,” International Robotics & Automation Journal, vol. 1, pp. 1–6, 2016.
14. T. Aladwani, “Impact of selecting virtual machine with least load on tasks schedul-
ing algorithms in cloud computing,” in Proceedings of the 2nd International Confer-
ence on Big Data, Cloud and Applications, 2017, pp. 1-7.
15. S. Elmougy, S. Sarhan, and M. Joundy, “A novel hybrid of Shortest job first and
round Robin with dynamic variable quantum time task scheduling technique,” Jour-
nal of Cloud computing, vol. 6, pp. 1–12, 2017.
16. R.K. Naha and M. Othman, “Cost-aware service brokering and performance sentient
load balancing algorithms in the cloud,” Journal of Network and Computer Applica-
tions, vol. 75, pp. 47–57, 2016.
17. K. Manojkumar and P. Kanagaraju, Enhanced load balancing algorithm to reduce
response time and waiting time by incorporating weighted round robin and honey
bee behaviour algorithm in cloud computing.
18. M.B. Rasheed, N. Javaid, M.S.A. Malik, M. Asif, M.K. Hanif, and M.H. Chaudary,
“Intelligent multi-agent based multilayered control system for opportunistic load
scheduling in smart buildings,” IEEE Access, vol. 7, pp. 23990–24006, 2019.
19. H. Zhong, Y. Fang, and J. Cui, “Reprint of “LBBSRT: An efficient SDN load
balancing scheme based on server response time,” Future Generation Computer
Systems, vol. 80, pp. 409–416, 2018.
20. S. Dam, G. Mandal, K. Dasgupta, and P. Dutta, “An ant-colony-based meta-heuristic
approach for load balancing in cloud computing,” in Applied Computational Intelli-
gence and Soft Computing in Engineering, IGI Global, 2018, pp. 204–232.
21. M. Kumar and S. Sharma, “Deadline constrained based dynamic load balancing al-
gorithm with elasticity in cloud environment,” Computers & Electrical Engineering,
vol. 69, pp. 395–411, 2018.
22. V. Bhavya, K. Rejina, and A. Mahesh, “An Intensification of Honey Bee Foraging
Load Balancing Algorithm in Cloud Computing,” International Journal of Pure and
Applied Mathematics, vol. 114, pp. 127–136, 2017.
23. N.X. Phi, C.T. Tin, L.N.K. Thu, and T.C. Hung, “Proposed load balancing algorithm
to reduce response time and processing time on cloud computing,” Int. J. Comput.
Networks Commun., vol. 10, pp. 87–98, 2018.
24. K. Sekaran, M.S. Khan, R. Patan, A.H. Gandomi, P.V. Krishna, and S. Kallam, “Im-
proving the response time of m-learning and cloud computing environments using a
dominant firefly approach,” IEEE Access, vol. 7, pp. 30203, 2019.
25. L. Xingjun, S. Zhiwei, C. Hongpong, and B.O. Mohammed, “ A new fuzzy-based
method for load balancing in the cloud based Internet of thing usingagrey wolf optiiza-
tion algorithm,’’ International Journal of Communication Systems, vol. 33, p. e4370,
2020.
A comprehensive survey on load balancing techniques for virtual machines
Системні дослідження та інформаційні технології, 2023, № 4 147
26. G. Natesan and A. Chokkalingam, “Task scheduling in heterogeneous cloud envi-
ronment using mean grey wolf optimization algorithm,” ICT Express, vol. 5,
pp. 110–114, 2019.
27. M.K. Halili and B. Cico, “SLA management for comprehensive virtual machine mi-
gration considering scheduling and load balancing algorithm in cloud data centers,”
International Journal on Information Technologies & Security, vol. 12, 2020.
28. A.K. Kiani and N. Ansari, “On the fundamental energy trade-offs of geographical
load balancing,” IEEE Communications Magazine, vol. 55, pp. 170–175, 2017.
29. N.H. Shahapure and P. Jayarekha, “Virtual machine migration based load balancing
for resource management and scalability in cloud environment,” International Jour-
nal of Information Technology, pp. 1–12, 2018.
30. X. Shao, M. Jibiki, Y. Teranishi, and N. Nishinaga, “An efficient load-balancing
mechanism for heterogeneous range-queriable cloud storage,” Future Generation
Computer Systems, vol. 78, pp. 920–930, 2018.
31. S. Subalakshmi and N. Malarvizhi, “Enhanced hybrid approach for load balancing
algorithms in cloud computing,” International Journal of Scientific Research in
Computer Science, Engineering and Information Technology, vol. 2, pp. 136–142, 2017.
32. T.G. Rodrigues, K. Suto, H. Nishiyama, N. Kato, and K. Temma, “Cloudlets activa-
tion scheme for scalable mobile edge computing with transmission power control
and virtual machine migration,” IEEE Transactions on Computers, vol. 67,
pp. 1287–1300, 2018.
33. M. Xu, W. Tian, and R. Buyya, “A survey on load balancing algorithms for virtual
machines placement in cloud computing,” Concurrency and Computation: Practice
and Experience, vol. 29, p. e4123, 2017.
Received 21.11.2022
INFORMATION ON THE ARTICLE
Suman Sansanwal, ORCID: 0000-0001-8485-0931, Chandigarh University, Punjab, In-
dia, e-mail: sumanphd27@gmail.com
Nitin Jain, Chandigarh University, Punjab, India, e-mail: nitin@gmail.com
КОМПЛЕКСНИЙ ОГЛЯД ТЕХНІК БАЛАНСУВАННЯ НАВАНТАЖЕННЯ
ДЛЯ ВІРТУАЛЬНИХ МАШИН / Суман Сансанвал, Нітін Джайн
Анотація. Хмарні обчислення — це нова техніка з чудовими характеристика-
ми, такими як масштабованість, висока гнучкість і надійність. Оскільки ця
сфера експоненціально зростає, швидке та якісне обслуговування приваблює
більше користувачів. Розподіл віртуальної машини (VM) відіграє вирішальну
роль в оптимізації хмарних обчислень; на розподіл ресурсів не впливає збій
машини та перенесення відбувається без простоїв. Ефективне керування вірту-
альними машинами необхідне для збільшення прибутку, енергозбереження
тощо. Однак воно може більш ефективно використовувати ресурси віртуальної
машини через збільшення навантаження, тому балансування навантаження є
більш концентрованим. Переважна мета балансування навантаження — рівно-
мірно збалансувати доступне навантаження між вузлами, щоб уникнути про-
блем із перевантаженням або недовантаженням. У дослідженні виконано роз-
ширений огляд щодо розміщення віртуальних машин, щоб описати
застосування алгоритмів прогнозування та забезпечити більш ефективне, на-
дійне розміщення віртуальної машини з високою відповіддю та низькими на-
кладними витратами. Крім того, у ході роботи зроблено спробу оглянути пробле-
ми балансування навантаження у розміщення віртуальної машини, а також
різні ідеї щодо сучасних методів вирішення цих проблем.
Ключові слова: розподіл віртуальної машини, балансування навантаження,
хмарні обчислення, перевантаження, фізична машина, центр оброблення даних.
|
| id | journaliasakpiua-article-267280 |
| institution | System research and information technologies |
| keywords_txt_mv | keywords |
| language | English |
| last_indexed | 2025-07-17T10:28:00Z |
| publishDate | 2023 |
| publisher | The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" |
| record_format | ojs |
| resource_txt_mv | journaliasakpiua/a8/902c77f3e8462b3a5d81a7ac0415cea8.pdf |
| spelling | journaliasakpiua-article-2672802024-02-01T21:03:07Z A comprehensive survey on load balancing techniques for virtual machines Комплексний огляд технік балансування навантаження для віртуальних машин Sansanwal, Suman Jain, Nitin розподіл віртуальної машини балансування навантаження хмарні обчислення перевантаження фізична машина центр оброблення даних virtual machine allocation load balancing cloud computing overloading physical machine data center Cloud computing is an emerging technique with remarkable features such as scalability, high flexibility, and reliability. Since this field is growing exponentially, more users are attracted to fast and better service. Virtual Machine (VM) allocation plays a crucial role in cloud computing optimization; hence, resource distribution is not impacted by machine failure and is migrated with no downtime. Therefore, effective management of virtual machines is necessary for increasing profit, energy-saving, etc. However, it could utilize the virtual machine resources more efficiently because of the increased load, so load balancing is more concentrated. The predominant purpose of load balancing is to balance the available load equally among the nodes to avoid overloading or underloading problems. The present study conducted an extensive survey on virtual machine placement to describe the application of prediction algorithms and to provide more efficient, reliable, high response, and low overhead VM placement. Furthermore, the survey attempted to overview the challenges in load balancing in VM placement and various ideas of state-of-the-art techniques to resolve the issues. Хмарні обчислення — це нова техніка з чудовими характеристиками, такими як масштабованість, висока гнучкість і надійність. Оскільки ця сфера експоненціально зростає, швидке та якісне обслуговування приваблює більше користувачів. Розподіл віртуальної машини (VM) відіграє вирішальну роль в оптимізації хмарних обчислень; на розподіл ресурсів не впливає збій машини та перенесення відбувається без простоїв. Ефективне керування віртуальними машинами необхідне для збільшення прибутку, енергозбереження тощо. Однак воно може більш ефективно використовувати ресурси віртуальної машини через збільшення навантаження, тому балансування навантаження є більш концентрованим. Переважна мета балансування навантаження — рівномірно збалансувати доступне навантаження між вузлами, щоб уникнути проблем із перевантаженням або недовантаженням. У дослідженні виконано розширений огляд щодо розміщення віртуальних машин, щоб описати застосування алгоритмів прогнозування та забезпечити більш ефективне, надійне розміщення віртуальної машини з високою відповіддю та низькими накладними витратами. Крім того, у ході роботи зроблено спробу оглянути проблеми балансування навантаження у розміщення віртуальної машини, а також різні ідеї щодо сучасних методів вирішення цих проблем. The National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute" 2023-12-26 Article Article Peer-reviewed Article application/pdf https://journal.iasa.kpi.ua/article/view/267280 10.20535/SRIT.2308-8893.2023.4.10 System research and information technologies; No. 4 (2023); 135-147 Системные исследования и информационные технологии; № 4 (2023); 135-147 Системні дослідження та інформаційні технології; № 4 (2023); 135-147 2308-8893 1681-6048 en https://journal.iasa.kpi.ua/article/view/267280/290679 |
| spellingShingle | розподіл віртуальної машини балансування навантаження хмарні обчислення перевантаження фізична машина центр оброблення даних Sansanwal, Suman Jain, Nitin Комплексний огляд технік балансування навантаження для віртуальних машин |
| title | Комплексний огляд технік балансування навантаження для віртуальних машин |
| title_alt | A comprehensive survey on load balancing techniques for virtual machines |
| title_full | Комплексний огляд технік балансування навантаження для віртуальних машин |
| title_fullStr | Комплексний огляд технік балансування навантаження для віртуальних машин |
| title_full_unstemmed | Комплексний огляд технік балансування навантаження для віртуальних машин |
| title_short | Комплексний огляд технік балансування навантаження для віртуальних машин |
| title_sort | комплексний огляд технік балансування навантаження для віртуальних машин |
| topic | розподіл віртуальної машини балансування навантаження хмарні обчислення перевантаження фізична машина центр оброблення даних |
| topic_facet | розподіл віртуальної машини балансування навантаження хмарні обчислення перевантаження фізична машина центр оброблення даних virtual machine allocation load balancing cloud computing overloading physical machine data center |
| url | https://journal.iasa.kpi.ua/article/view/267280 |
| work_keys_str_mv | AT sansanwalsuman acomprehensivesurveyonloadbalancingtechniquesforvirtualmachines AT jainnitin acomprehensivesurveyonloadbalancingtechniquesforvirtualmachines AT sansanwalsuman kompleksnijoglâdtehníkbalansuvannânavantažennâdlâvírtualʹnihmašin AT jainnitin kompleksnijoglâdtehníkbalansuvannânavantažennâdlâvírtualʹnihmašin AT sansanwalsuman comprehensivesurveyonloadbalancingtechniquesforvirtualmachines AT jainnitin comprehensivesurveyonloadbalancingtechniquesforvirtualmachines |