A Study of Load Balancing in Cloud Computing

This study focuses on load balancing in cloud computing, which involves the even distribution of resources to optimize performance. It explores different load balancing algorithms and their impact on resource utilization and job response time. The research aims to develop a dynamic load balancing algorithm and analyze its effectiveness in improving system efficiency and reducing machine overheads. The study also examines the challenges and issues associated with load balancing in hybrid cloud environments.

Running head: A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
A Study of Load Balancing in Cloud Computing
Name of the Student
Name of the University

2A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Abstract
The load balancing of a machine is performed dynamically by shifting the workload to the
remote machines or nodes from the local machines. In other words, the algorithm for load
balancing shifts the workload or traffic or the incoming server requests from the over utilized
servers to the underutilized ones. The fundamental purpose of load balancing is to ensure an
even distribution of the cloud computing resources in the hybrid cloud environments. The
fundamental function of the simplest kind is a cloud user establishes a connection with the
cloud with the help of a cloud broker or a cloud service provider. The cloud user is
responsible for submitting the request to the cloud for the required resource through the cloud
service provider. The study focuses on developing a load balancing algorithm for a virtual
machine to increase the system efficiency, performance, throughput and scalability. The
scope of the study encompasses a thorough and detailed understanding and practical
knowledge of the dynamic load balancing techniques in a hybrid cloud environment.

3A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Acknowledgement
Conducting this research has been one of the most enriching experiences of my life. The
contribution of this research to enhance my knowledge base and analytical skill has been
paramount. It gave me the opportunity to face challenges in the process and overcome them.
This would not have been possible without the valuable guidance of my professors, peers and
all the people who have contributed to this enriching experience. I would like to take this
opportunity to thank my supervisor ____________________________________ for the
constant guidance and support provided to me during the process of this research. It would
not be justified if I did not thank my academic guides for their important and valuable
assistance and encouragement throughout the research process. I would also like to thank my
friends who had provided me with help and encouragement for collecting primary data and
valuable resources. Finally, I would like to thank the professionals from the retail industry
who have participated in the research survey and provided with valuable inputs into the
subject. The support of all these people has been inspiring and enlightening throughout the
process of research in the subject.
Heartfelt thanks and warmest wishes,
Yours Sincerely,

4A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Table of Contents
Abstract......................................................................................................................................2
1. Introduction............................................................................................................................6
1.1 Background of the Study..................................................................................................6
1.2 Purpose of the Study........................................................................................................7
1.3 Research Aim...................................................................................................................7
1.4 Research Objectives.........................................................................................................7
1.5 Research Questions..........................................................................................................8
1.6 Problem Statement...........................................................................................................8
1.7 Structure of the Thesis.....................................................................................................9
2. Literature Review.................................................................................................................11
2.1 Issues and challenges in hybrid clouds..........................................................................12
2.2 Load balancing strategies in the cloud...........................................................................13
2.2.1 Static load balancing algorithms.............................................................................13
2.2.2 Dynamic load balancing algorithms........................................................................14
2.3 The need for load balancing...........................................................................................15
2.4 Load balancing algorithms.............................................................................................16
2.5 Policies and strategies for dynamic load balancing.......................................................18
2.6 Qualitative metrics for load balancing:..........................................................................19
2.7 Load balancing with cost scheduling algorithm:...........................................................20
2.8 Proposed system.............................................................................................................21
2.9 Design of the proposed algorithm..................................................................................22

5A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
2.10 Load Balancing Strategies in Hybrid Clouds...............................................................24
3. Research Methodology.........................................................................................................25
3.1 Methodology Selection..................................................................................................25
3.2 Method Outline..............................................................................................................25
3.3 Load Balancing Method.................................................................................................26
3.3 Research Planning..........................................................................................................26
4. Experiments & Results.........................................................................................................28
4.1 Working of Elastic Load Balancing...............................................................................29
4.2 Routing Algorithm.........................................................................................................31
4.3 Authentication and access control for load balancers....................................................31
} 4.4 API Actions with No Support for Resource-Level Permissions................................32
4.5 Resource level permissions for elastic load balancing...................................................32
4.6 Condition Keys for elastic load balancing.....................................................................35
5. Conclusion............................................................................................................................40
5.1 Limitation of the Study..................................................................................................40
5.2 Future Scope of the Study..............................................................................................41
References................................................................................................................................42

6A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Table of Figures and Tables
Figure 2.1: Classification of the load balancing algorithms........................................16
Figure 2.2: The basic flow of execution of the load balancing algorithm...................19
Table 2.1: Key variables used in load balancing..........................................................22
Figure 2.3: Proposed system for load balancing in loud..............................................23
Figure 2.4: Flowchart of the load balancing algorithm................................................25
Table 3.1: Research timetable......................................................................................28

7A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
1. Introduction
1.1 Background of the Study
Cloud computing is seen to be originated or a combination of the parallel computing
grid computing and the distributed computing. It is seen that most of computers as well as the
storage devices are connected to are connected and form a heterogeneous pool of resources.
Additionally the virtualization technology is associated with the utilization of the different
resources by means of providing various kind of resources in the internet, which can be
brought or paid in accordance to the demands of the user. Cloud computing is associate with
providing of different kind of services which includes SaaS or the Software as a Service,
PaaS, Platform as a Service, IaaS or infrastructure as a Service (Bhatt and Bheda, 2017). In
order to utilize the resources in a proper way on the virtual machines and the physical
machine load balancers there exists need of distribution of the resources. Load Balancing
generally refers to the method of optimizing the resources in the clouds virtual machines and
the main purpose of the load balancers includes the distribution of the resources as well as the
tasks amongst all the machines in an equal manner so as to make sure that none of the nodes
are overloaded or is idle. Some of the major reasons lying behind the usage of the load
balancing includes the following:
 Reduction is the time of waiting
 Reduction of the response time
 Increased utilization of the resources
 Identification of the reliability
 Increasing the through output
 Load balancing is associated with enhancing the performance of the entire
system by management of each and every node

8A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
1.2 Purpose of the Study
Despite of the existence of the various dynamic and the static algorithm which have
been proposed and developed in recent time in which each of the algorithm is having its own
advantages as well as disadvantages. The research is mainly focused upon the solving of the
execution time of the nodes along with the overloading upon the various machines. The
research would also be associated with finding a solution for the waiting time of the high
time-consuming tasks along with helping in decreasing the delays in the operations (Kumar
and Kalra, 2016). The entire study is associated with focusing upon the different load
balancing algorithm along with finding of solutions for the different types of issues faced by
the load balancing.
1.3 Research Aim
The aim of the research is to look at even distribution of the cloud computing
resources in the hybrid cloud environments. It is also referred to as ‘load balancing’. Load
balancing is a useful mechanism that is typically useful in terms of ensuring proper utilization
of resources as well as job response time. As a result, load balancing is responsible for
ensuring better performance results. On the other hand, proper load balancing techniques can
effectively reduce the amount of energy consumption and carbon footprint/ carbon dioxide
emission by means of evenly distributing the workloads. Therefore, this paper is dedicated to
study the various load balancing techniques and algorithms in order to compare and contrast
between the advantages and disadvantages associated with the most commonly used load
balancing algorithms.
1.4 Research Objectives
The main objectives of the entire study have been listed below:
 To develop a dynamic load balancing algorithm,

9A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
 To monitor the virtual machine
 To manage the virtual resources in an efficient manner
 To increase the throughput of the system
 To reduce the machine overheads
 To analyse the response time, waiting time, performance, power, speed,
scalability, power consumption of the virtual machines.
1.5 Research Questions
In accordance with the research objectives, the research questions are formulated as
follows:
 How to develop a dynamic load balancing algorithm?
 How to monitor the virtual machine?
 How to manage the virtual resources in an efficient manner?
 How to increase the throughput of the system?
 How to reduce the machine overheads?
 How to analyse the response time, waiting time, performance, power, speed,
scalability, power consumption of the virtual machines?
1.6 Problem Statement
Virtualization is associated with acting as the backbone of the cloud computing and
the virtualization is the concept which is associated with including the computing processing
along with the storage, the assigning of the tasks. In addition to this, the research is associated
with helping in managing of the resources of the physical devices so as to create different
virtual machines. This in turn, is associated with putting forward of the problem that how
much resources are to be assigned to the machines and what are the resources which are
needed by that machine after the resources are allocated to the machines. This in turn is

10A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
associated with giving rise to another problem that is what tasks are to be performed by
which machine or what task is to be assigned to which machine. For this reason, the load
balancing is used so as to solve all this problems. This is a very important task that the virtual
machines are having along with the task management.
1.7 Structure of the Thesis
The research paper will be segregated into several individual or separate chapters or
sections, each of which will typically focus on the different aspects and areas of the topic of
load balancing in cloud computing. The main content of these chapters will be as follows:
Introduction: The introduction chapter will focus on the basic idea or concept of the
study topic i.e. load balancing in cloud computing. Apart from that, it will also involve the
research aim and objectives along with an explanation of the basic rationale or purpose of the
research.
Literature review: The literature review section thoroughly describes the previously
existing studies and theories that have been undertaken in this research area. To be more
precise, the crucial review and analysis of existing literature on cloud computing and load
balancing involves studying peer reviewed journals, research articles, thesis papers, research
papers, books, blogs and others.
Research methodology: The research methodology section describes the appropriate
or most suitable approach that will be undertaken in order to carry out the research in the
desired way and reach towards a successful conclusion.
Experiments and results: This section critically examines the results and outcomes
yielded from the research as carried out in this study along with the thorough and detailed
description of the outcomes.

11A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Conclusion: The conclusion chapter briefly gives an overview of the overall research
as undertaken in this study and portrays the relevant and useful outcomes as discussed in the
previous chapters.

12A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
2. Literature Review
Xen is a bare metal of type-1 and it is an open source virtual machine monitor (VMM)
that is used in commercial clouds, such as Amazon EC2 and Rackspace. It has the
characteristics of both hardware and software. The virtual machines, which are known as a
domain in Xen, contain a number of VCPU’s that are schedules by a VMM scheduler on a
host by using multiple physical CPUs. Xen creates a domain called domain 0 that manages
the guest domains. Fixed amount of time based on Round Ribbon is assigned to each task and
a ring is used to arrange the tasks as time is equally assigned to each node. The task is
assigned to the virtual machine by the data centre controller in the form of a ring picked from
the VM in a random manner. The problem, which it faces, is that the state of VM is not
checked. The throttled load balancing algorithm has the list of VM along with their states
when required by clients the list of virtual machines are scanned and then task is assigned.
The tasks, which require less resources and complication time, are worked upon by
Min-Min Scheduling algorithm. In case of Min-Max, scheduling algorithm the load balancer
picks up the task that needs more resources and completes the task taking some time. The
task completion time is too long in Min-Max algorithm. Ant colony load balancing algorithm
is an algorithm in which when a task is received, the ant moves from these source to the super
node. The records of the nodes, which have been visited or have been assigned any task, are
kept. Honeybee load balancing algorithm works in a heterogeneous environment. The
machines are grouped based on their load after their states are recorded. The new task is
allocated to the machine that is least loaded.
Resource-aware scheduling algorithm is a mixture of Min-Min and Min-Max
algorithm and this algorithm takes up the advantages of both discarding the flaws. Active
monitoring load balancing algorithm is tested through simulation and the result of simulation

13A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
is shared. Least Connection Algorithm is based upon a connection on a single load. The load
balancer transfers the load to that node which has the least number of connections. The load
balancer handles all the information related to increasing and decreasing connection.
2.1 Issues and challenges in hybrid clouds
There are multiple challenges and issues associated with the hybrid cloud
environments. These are discussed briefly as follows:
Portability and interoperability: In terms of outsourcing, vendor lock in is one of the
most important aspects. In this context, the problem of interoperability and portability is
discussed openly in hybrid clouds by various researchers. Interoperability refers to the way
communication between the multiple clouds take place. For example, Google and Amazon
both uses the image of Windows without any change. Therefore, it can be considered as an
example of interoperability. On the other hand, portability defines the capacity or capability
of the cloud in terms of moving application and data from one cloud to the other. If the
dependencies are removed, it can be considered as a potential possibility.
Denial of service: It is another potential challenge associated with cloud computing
environments as identified by the researchers. According to Mesbahi and Rahmani (2016),
the hybrid cloud and even the normal cloud environments involve resources that are typically
allocated in a dynamic manner. In addition to that, it is crucial to understand how these cloud
computing environments respond to a denial of service (DoS) attack. In hybrid clouds, when
resources are unavailable for executing a particular task, then those particular tasks are
forwarded to the public clouds. However, in this case, the discussed strategy is not a feasible
solution. Therefore, to the researchers, it is a burning challenge with respect to cloud
computing.

14A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Load balancing: Another crucial challenge faced in hybrid clouds is associated with
load balancing. It is associated with the need to distribute workloads in an even and dynamic
fashion between the nodes in the public and private clouds. In distributed systems, the term
‘load balancing’ is typically referred to the process of distributing the workloads between the
various nodes in order to enhance the overall job response time and utilization of resources.
While this distribution process is going on, it is essential to ensure that none of the nodes are
heavily loaded or assigned tasks lesser than their capacity or left idle. Moreover, it is essential
to make sure that all the nodes are assigned an equal amount of load individually.
In this context, Kumar and Kumar (2019) suggests that if resource utilization is
performed in an optimal way, then the system performance will increase automatically. In
addition to that, the degree of carbon emission and amount of energy consumption is also
subjected to reduce effectively. Furthermore, Devi and Uthariaraj (2016) pointed out that the
possibility of bottleneck due to imbalance of loads can also effectively reduce. Therefore, the
even distribution of loads between the nodes enables a fair and efficient resource utilization,
which in turn helps in greening of those concerned environments.
2.2 Load balancing strategies in the cloud
Load balancing algorithms can be categorized into different classes for better
understanding purposes. Classifying the load balancing algorithms can potentially help in the
identification of appropriate algorithm based in the time of need. In this context, Adhikari and
Amgoth (2018) has said that the emerging concept of hybrid clouds are increasingly
associated with the idea of balancing the load between the private and public clouds. Several
studies have been conducted in this area, as this phenomenon is given prime importance in
the hybrid cloud environment. In the recent past, researchers have also tried to balance the
loaf at the peak times at the time of meeting the predefined SLAs (Service Level
Agreements) and QoS (Quality of Service).

15A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
2.2.1 Static load balancing algorithms
According to Chen, Chen and Kuo (2017), the cloud environments involve huge
amount of workloads on load balancing in homogeneous resources. On the other hand,
several research on heterogeneous environment and the related load balancing has also
received significant spot light. In this context, the past researchers have already studied the
effect of round robin technique following a dynamic approach depending on variable cloudlet
long length, host bandwidth, virtual machine bandwidth and virtual machine (VM) image size
(Gandhi et al., 2015). The load optimization is performed by means of adjusting these
varying parameters. For this implementation, the CloudSim simulator model is used.
2.2.2 Dynamic load balancing algorithms
Batrouni, Zions and Homiou (2017) had presented a hybrid load balancing policy in
the hybrid cloud environments. This particular policy consists of two individual stages such
as the (i) Dynamic load balancing stage or (ii) static load balancing stage. The stages are
focused on the selection of a suitable node set in the static load balancing stage. In other
words, the dynamic load balancing algorithm uses the current state of the system for making
the decisions for load balancing. Hence, it shifts the load depending on the current state of the
system. It further allows the processing to move from the over utilized machines to the under
utilized machines in a dynamic manner to ensure faster execution of the workloads (Farrag,
Mahmoud and El Sayed, 2015). As a result, it typically allows for process preemption, that
static load balancing approach does not support. A crucial advantage of this particular
approach is that its decision in terms of balancing the load is solely depending on the current
state of the system. Therefore, it significantly helps to improve the overall performance of the
systems by means of dynamic migration of the load.

16A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Figure 2.1: Classification of the load balancing algorithms
(Source: Kapoor and Dabas, 2015)
2.3 The need for load balancing
The load balancing of a machine is performed dynamically by shifting the workload
to the remote machines or nodes from the local machines (i.e. to the under utilized nodes
from the over utilized ones). As a result, this approach to load balancing is typically useful
maximizing the user satisfaction and minimizing the response time. It further increases the
utilization of resources and reduces the number of job rejections. Moreover, it also raises the
performance ratio of the concerned system. In addition to that, load balancing is also effective
in terms of achieving the green computing paradigm in the clouds. The specific factors that
are responsible for load balancing through green computing are as follows:
Reduction of carbon emission: The consumption of energy and emission of carbon
dioxide are the two different sides of the same coin. In other words, these two paradigms are
closely interconnected. According to Gopinath and Vasudevan (2015), they are directly
proportional to each other. The concept of load balancing is useful in terms of reducing the

17A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
amount of energy consumption that will automatically enable potential reduction of the
amount of carbon emission and thereby, achieve Green Computing.
Limited energy consumption: The concept of load balancing can effectively reduce
the degree of energy consumption by means of avoiding the process of modes or VM (virtual
machine) over heating due to the massive or excessive workloads.
2.4 Load balancing algorithms
There are three major algorithms for load balancing. These are equally spread current
execution load, round robin, Adaptive Resource Allocation (ARA) and throttled load
balancing.
Equally spread current execution load: The specific algorithm needs a typical load
balancer for monitoring the jobs for asked for execution. The load balancer has to queue up
the jobs for handing over to the different VMs or virtual machines. The load balancer further
supervises the said queue in a frequent basis for new jobs and then performs job allocations to
the list of the free VMs or virtual servers (Chien, Son and Loc, 2016). The balance is also
responsible for maintaining the list of tasks as allocated to the virtual servers, it helps in the
identification of the virtual machines to be able to allocate them with the new jobs. According
to Milani and Navimipour (2016), for this algorithm, the experimental work is typically
performed with the help of a cloud analyst simulation. As the name suggests, this present
algorithm only operates on the equally spread execution load on the various kinds of virtual
machines or VMs.
Round robin: This particular load balancing algorithm typically uses the time scaling
mechanism. The name ‘round robin’ suggests that this algorithm works in a round fashion
where the nodes are allocated with a particular time slice. Each node is subjected to wait for
its turn. On the other hand, the total time is divided into multiple slices and each of the nodes

18A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
are allocated the intervals. The complicity of this particular algorithm is significantly lower in
comparison with the other two algorithms. In this context, Singh, Juneja and Malhotra (2015)
has suggested that an open source simulation is required to be performed for the algorithm
software to recognize as a cloud analyst. This algorithm is default for simulation purposes. In
addition to that, the round robin algorithm simply allots a specific job in a round manner or
fashion that does not take into consideration the load on the different virtual machines.
ARA (Adaptive Resource Allocation): In Adaptive Resource Allocation (ARA)
algorithm executed in the cloud systems, it attempts for counteracting the deleterious impact
of burrstones by means of providing some typical randomness in the process of decision-
making. Therefore, it effectively improves the overall system availability and performance.
The main problem with this particular strategy is it solely takes into consideration the Poisson
arrival streams, the service time (exponentially distributed), and the fixed number of choice.
Throttled load balancing: This algorithm works by means of finding the most
suitable or appropriate VM (virtual machine) in order to assign a specific job. The job
manager is typically the owner of a list of all the VMs. It uses the indexed list for allotting a
desired job to the appropriate virtual machine. If the job is well suited for a specific virtual
machine, then the job is to be assigned to the appropriate VM machine. On the other hand,
the virtual machine is available in terms of accepting the jobs. At the time, the job manager
waits for the specific client request and takes the job from the concerned queue for processing
it in a faster and efficient way.
The figure below demonstrates the diagrammatical representation of the load
balancing algorithms as used for balancing load in the cloud computing environment.

19A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Figure 2.2: The basic flow of execution of the load balancing algorithm
(Source: Paya and Marinescu, 2015)
2.5 Policies and strategies for dynamic load balancing
The different policies and strategies for dynamic load balancing are discussed as
follows:
Transfer policy: The transfer policy is typically used to select a process or a task from
a local machine in order to transfer it to a remote virtual machine.
Location policy: The location policy is specifically used by a machine or a processor
in order to share the transferred task with the help of an overloaded machine.
Selection policy: The selection policy identifies the machines or processors, which
typically takes part or participates in the load balancing activity.
Information policy: The information policy accounts for collecting the useful data
based on which the load balancing decisions are taken.

20A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Process transfer policy: The process transfer policy is useful in order to decide the
execution of a task (to be done either remotely or locally).
Load estimation policy: The load estimation policy decides the method used in order
to approximate the entire amount of workload of a machine or a processor.
Migration limiting policy: The migration policy us effective in terms of setting a
boundary to the maximum number of times that a particular task can migrate from one
processor or machine to the other.
Priority assignment policy: The priority assignment policy is responsible for assigned
the appropriate priority in order to execute both the remote and the local tasks and processes.
2.6 Qualitative metrics for load balancing:
The qualitative parameters or metrics are crucial for performing successful load
balancing in a typical hybrid cloud environment. These qualitative metrics are discussed as
follows:
Fault tolerant: Fault tolerant load balancing refers to the ability to execute the
algorithm in a uniform and correct manner so that the certain difficult conditions (for
example, failure in an arbitrary node in the system) can be dealt efficiently.
Throughput: Throughput refers to the total number of tasks, which achieved
successful completion in terms of execution. It is essential to maintain high throughput for
the system to perform in a better way.
Associated overhead: Associated overhead refers to the exact amount of overhead
being produced while a particular load balancing algorithm is executed.
Performance: The performance of a load balancing denotes the degree of
effectiveness in the overall system after the load balancing has been performed.

21A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Scalability: Load balancing algorithm depends on the system’s ability to support
scalable operations. A highly scalable load balancing is more effective as it works well with a
restricted number of machines or processors (Ghomi, Rahmani and Qader, 2017).
Response time: Response time refers to the minimum amount of time taken by a
distributed system in terms of executing a particular load balancing algorithm in order to
respond.
Migration time: The total time undertaken to transfer or migrate a task from one
machine to another machine is referred to as the migration time. In order to improve the
performance level of a particular system, the migration time should be minimized.
Resource utilization: Resource utilization essentially indicates the degree or extent up
to which the system utilizes the resources. Maximum resource utilization is ensured in a
proper load balancing algorithm.
2.7 Load balancing with cost scheduling algorithm:
The fundamental function of the simplest kind is a cloud user establishes a connection
with the cloud with the help of a cloud broker or a cloud service provider. The cloud user is
responsible for submitting the request to the cloud for the required resource through the cloud
service provider. On the other hand, the cloud service provider is responsible for ensuring
optimal effectiveness and efficiency. In order to provide better service to the cloud user, it is
essential to apply the optimization algorithm (Dave, Patel and Bhatt, 2016). The request
execution actually takes place in the cloud using the virtual machines and at the same time,
the deployment of the available resource pool. These are typically available as cloud
middleware. These available resources are served as storage devices, network device or
operating system device. In addition to that, it also demonstrates the way load balancer is
responsible for distributing the load among the different virtual servers or VMs so that the

22A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
entire processing of the resource request is achieves successful completion and provisioning
of the service to the user. Moreover, the load processing distribution is done amongst
multiple virtual machines by making sure that none of the VMs get overloaded. The notation
as represented in the table below are used for defining the key variables. The cost is defined
with C. The execution cost essentially depend on the package that contains the resource on a
specific virtual machine.
Variables Meaning
Ri The available cloud resources
VMi The available VMs or virtual machines
Ci The fixed price for VMi executing Ri
u_cost The cost of the user for getting the service
e_cost The cost taken by the virtual machine to
serve the user
u_time Waiting time of the user
Pri Profit at provider for executing the resource
I Number of instances that ranges from 1 to n
P Processor
Pkg Resource grouped into packages
Table 2.1: Key variables used in load balancing
(Source: Xu, Tian and Buyya, 2017)
2.8 Proposed system
In accordance with the proposed architecture, the requests are coming from multiple
users where every request is containing different or dissimilar task. At each task, a different

23A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
parameter is measured for price, importance, processor request and time. Time takes into
consideration the time required for computation, as required for completing the task. The
cloud administrator monitor the new and old cloud customers as well as charges the price
parameters.
Figure 2.3: Proposed system for load balancing in loud
(Source: Oueis, Strinati and Barbarossa, 2015)
2.9 Design of the proposed algorithm
Step 1: The first step deals with the initiation of the data center (DC) broker. The state
of the existing clouds and the status table of the VM (virtual machine) is considered.
Step 2: When a request is received for allocating a new VM (virtual machine) to a
data center broker, it is essential for the DC broker to analyze the status table. After that, the
calculation of the total amount of time required for execution of all the existing cloudlets in
the queue with respect to the VMs. The expected time for completion of the cloudlet
preparation for processing is based on the smallest processing time and the machine is
submitted in the next or following cloudlet. In case of multiple virtual machines, the first
virtual machine has been selected (Aslam and Shah, 2015).

24A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Step 3: The selected virtual machine is sent to the DC broker. Then, the data center
broker is responsible for sending the cloudlet to the VM, which is then allocated by the ID
number.
Step 4: The data center broker is responsible for notifying the updates and new
allocations to cloud status tables and virtual machine.
Step 5: After the completion of the processing request by the virtual machines, the
data center broker is responsible for receiving the Cloudlet response. Accordingly, it will be
updating the status table as completed along with the reduction of cloudlet in status table.
Step 6: Go to Step 2.

25A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Figure 2.4: Flowchart of the load balancing algorithm
(Source: Venkiteswaran and Shah, 2018)
2.10 Load Balancing Strategies in Hybrid Clouds
According to Zhao et al. (2015), the load balancing of a machine is performed
dynamically by shifting the workload to the remote machines or nodes from the local
machines (i.e. to the under utilized nodes from the over utilized ones). As a result, this
approach to load balancing is typically useful maximizing the user satisfaction and
minimizing the response time. In order to provide better service to the cloud user, it is
essential to apply the optimization algorithm. The request execution actually takes place in
the cloud using the virtual machines and at the same time, the deployment of the available
resource pool. These are typically available as cloud middleware.
Mousavi, Mosavi and Varkonyi-Koczy (2017) had presented a hybrid load balancing
policy in the hybrid cloud environments. This particular policy consists of two individual
stages such as the (i) Dynamic load balancing stage or (ii) static load balancing stage. The
stages are focused on the selection of a suitable node set in the static load balancing stage. In
other words, the dynamic load balancing algorithm uses the current state of the system for
making the decisions for load balancing. When a request is received for allocating a new VM
(virtual machine) to a data center broker, it is essential for the DC broker to analyze the status
table. After that, the calculation of the total amount of time required for execution of all the
existing cloudlets in the queue with respect to the VMs.

26A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
3. Research Methodology
3.1 Methodology Selection
The research would be conducted by means of collection of the secondary information
from the various sources such as journals, research articles and some media sources. The
previous research papers will be investigated for understand the findings of the previous
researchers and understanding the gaps. Both the two types of data will be qualitative in
nature. The primary data will be collected by interviewing some managers of the business
organizations that are using accounting information systems. The secondary data will be
collected from the secondary sources like peer reviewed journals, company’s annual reports,
news etc. Based on the previous findings the research study will be conducted for further
investigation towards usage of the load balancing in cloud computing.
3.2 Method Outline
The research project is experimental in nature, which means that the research project
will involve an experimental study of the preliminary load balancing technique in the hybrid
cloud based environment. In order to serve this purpose, the research will propose a basic
system of load balancing that includes the fundamental steps that are required to be
performed in order to ensure equal distribution of the load between the several virtual
machines or nodes. The experiments and results chapter describes the outcomes of the
analysis and experiments as has been undertaken by the research project. Apart from that, the
research project also involves an extensive review of secondary data and literature sources for
executing the study in a proper manner. The critical review and in depth study of the
previously existing research studies in the load balancing in cloud computing essentially
helps in gaining sufficient knowledge and information regarding the various aspects and areas

27A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
of the topic of concern. Therefore, an extensive secondary literature review is important in
order to execute the experimental study in a proper manner.
3.3 Load Balancing Method
There are several load balancing methods specifically designed to ensure efficient
distribution of incoming server requests or workloads or traffic among the virtual or physical
servers in the pool of servers. In this paper, an effective load balancing technique or method
will be developed with an aim to manage the virtual resources in an effective manner. In
addition to that, the project also aims to monitor the virtual machine. Besides, other
considerations involve the reduction of machine overheads, increase of the system throughput
and analysis of the performance, waiting time, migration time, response time and scalability
of the VMs (Virtual Machines).
3.3 Research Planning
Task Name Duration Start Finish
Resource
Names
Research Proposal 16 29/04/2019 14/05/2019
Background study 6 15/05/2019 17/05/2019
Online library
Team member 1
Data Collection 28 18/05/2019 14/06/2019
Team members
Persons
associated with
usage of the
Load balancing
Literature Review 14 15/06/2019 28/06/2019 Team member

28A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Analysis of the
Findings
16 29/07/2019 14/07/2019 Entire Team
Final Project
Submission
12 15/07/2019 26/07/2019
Table 3.1: Research timetable
(Source: Created by the learner)

29A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
4. Experiments & Results
Based upon the data and findings gathered from the literature review sections have
been focused on the finding out the results of the objectives initiated. The load balancer is
capable of handling the session persistence whenever needed. The research would also be
associated with finding a solution for the waiting time of the high time-consuming tasks
along with helping in decreasing the delays in the operations. The entire study is associated
with focusing upon the different load balancing algorithm along with finding of solutions for
the different types of issues faced by the load balancing. Amazon Web Services elastic load
balancer tool have been used for monitoring load on the cloud server. A load balancer helps
in distributing incoming traffic across targets including EC2 instances. This has helped in
enabling increase of availability of application. The load balancer has been used in
monitoring the health of registered targets and ensures that routing is done to traffics in
healthy targets. Configuring load balancer for accepting incoming traffic has been done with
listeners that are connected with a protocol and port number for configuring from clients to
load balancer. Elastic load balancing supports various types of load balancers including
application load balancers, network load balancers and classic load balancers. An application
load balancer makes routing and load balancing decisions at application layer
(HTTP/HTTPS). A network load balancer helps n routing and load balancing decisions at the
transport layer (TCP/TLS). Both application load balancers and network load balancers have
been able to route requests to one or more ports on each EC2 instance in virtual private cloud
(VPC). A Classic Load Balancer makes routing and load balancing decisions either at the
transport layer (TCP/SSL) or the application layer (HTTP/HTTPS), and supports either EC2-
Classic or a VPC. For more information.
For getting started, follow these steps:

30A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
 Create a load balancer using CreateLoadBalancer.
 Create a target group using CreateTargetGroup.
 Register targets for the target group using RegisterTargets.
 Create one or more listeners for your load balancer using CreateListener.
For deleting a load balancer and related resources, follow these steps:
 Delete the load balancer using DeleteLoadBalancer.
 Delete the target group using DeleteTargetGroup.
4.1 Working of Elastic Load Balancing
A load balancer helps in accepting incoming traffic from clients and routes requests to
registered targets in one or more Availability Zones. The load balancer has been helping in
monitoring health of its registered targets or ensuring routing traffic to healthy targets. In
case, load balancer detects an unhealthy target, it stops routing traffic to that target and
resume routing traffic to that target as it detects it’s as healthy.
When an availability zone has been enabled for load balancer, elastic load balancing
helps in creating load balancer node in availability zone. When register targets in an
availability zone and not able to enable availability zone, these registered targets don it
receive traffic. It has been recommended that multiple availability zones need to be enabled.
This help in case one availability zone is unavailable and has no healthy targets, the load
balancer still continue to route traffic to healthy targets on different availability zone. In case
availability zone, is disabled, the targets is that availability zone remain registered with load
balancer.
The nodes of load balancer distribute requests from clients for registering targets.
When cross zone load balancing has been enabled, each load balancer node helps in
distributing traffic across the registered targets in all enabled availability zones. In case cross

31A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
load balancing has been enabled, each load balancer nodes helps in distributing traffic across
the registered targets in various enabled availability targets. The following diagrams
demonstrate the effect of cross-zone load balancing. There are two enabled Availability
Zones, with 2 targets in Availability Zone A and 8 targets in Availability Zone B. Clients
send requests, and Amazon Route 53 responds to each request with the IP address of one of
the load balancer nodes. This distributes traffic such that each load balancer node receives
50% of the traffic from the clients. Each load balancer node distributes its share of the traffic
across the registered targets in its scope.
In case, cross zone load balancing has been disabled, each of the 2 targets in
availability zone A has been receiving 25% of the traffic and each of the 8 targets in
availability zone B receives 6.25% of the traffic. Each load balancer node has been routing
50% of the client traffic only towards target in availability zone.

32A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Network Load Balancers and cross zone load balancing has been disabled by default.
The default classic load balancer has been maintaining a keen approach in the dependency of
load balancing technology.
4.2 Routing Algorithm
The load balancer node has been receiving requests that have been evaluating listener
rules in priority order for determining which rule to be applied and selects target from the
target group for the rule action using the round algorithm. The load balancer node that has
been receiving connection used to select a target from the target group using flow hash
algorithm.
4.3 Authentication and access control for load balancers
AWS has been using security credentials that have been identifying and granting
access to AWS resources. Various features can be accessed using AWS identity and access
management (IAM) for allowing other users services and applications to your AWS resource
fully and sharing credentials.
An IAM policy has been a JSON document that include one or more statements. Each
statements has been structured as follows:
{

33A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
"Version": "2012-10-17",
"Statement":[{
"Effect": "effect",
"Action": "action",
"Resource": "resource-arn",
"Condition": {
"condition": {
"key":"value"
}
}
}]
}
4.4 API Actions with No Support for Resource-Level Permissions
Following are the elastic load balancing actions that not support resource-level permissions:
API version 2015-12-01:
• DescribeAccountLimits
• DescribeListenerCertificates
• DescribeListeners
• DescribeLoadBalancerAttributes
• DescribeLoadBalancers
• DescribeRules
• DescribeSSLPolicies
• DescribeTags
• DescribeTargetGroupAttributes
• DescribeTargetGroups
• DescribeTargetHealth
• API version 2012-06-01:
• DescribeInstanceHealth
• DescribeLoadBalancerAttributes
• DescribeLoadBalancerPolicyTypes

34A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
• DescribeLoadBalancers
• DescribeLoadBalancerPolicies
• DescribeTags
4.5 Resource level permissions for elastic load balancing
The following are the elastic load balancing that has been supporting resource level
permission and supported resource for each action:
API version 2015-12-01
API Action Resource ARNs
AddListenerCertificates listener
AddTags load balancer, target group
CreateListener load balancer
CreateLoadBalancer load balancer
CreateRule listener
CreateTargetGroup target group
DeleteListener listener
DeleteLoadBalancer load balancer
DeleteRule listener rule
DeleteTargetGroup target group
DeregisterTargets target group
ModifyListener listener
ModifyLoadBalancerAttributes load balancer
ModifyRule listener rule
ModifyTargetGroup target group
ModifyTargetGroupAttributes target group
RegisterTargets target group

35A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
RemoveListenerCertificates listener
RemoveTags load balancer, target group
SetIpAddressType load balancer
SetRulePriorities listener rule
SetSecurityGroups load balancer
SetSubnets load balancer
API version 2012-06-01
API Actions Resource ARNs
AddTags load balancer
ApplySecurityGroupsToLoadBalancer load balancer
AttachLoadBalancerToSubnets load balancer
ConfigureHealthCheck load balancer
CreateAppCookieStickinessPolicy load balancer
CreateLBCookieStickinessPolicy load balancer
CreateLoadBalancer load balancer
CreateLoadBalancerListeners load balancer
CreateLoadBalancerPolicy load balancer
DeleteLoadBalancer load balancer
DeleteLoadBalancerListeners load balancer
DeleteLoadBalancerPolicy load balancer
DeregisterInstancesFromLoadBalancer load balancer
DetachLoadBalancerFromSubnets load balancer
DisableAvailabilityZonesForLoadBalancer load balancer

36A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
EnableAvailabilityZonesForLoadBalancer load balancer
ModifyLoadBalancerAttributes load balancer
RegisterInstancesWithLoadBalancer load balancer
RemoveTags load balancer
SetLoadBalancerListenerSSLCertificate load balancer
SetLoadBalancerPoliciesForBackendServer load balancer
SetLoadBalancerPoliciesOfListener load balancer
4.6 Condition Keys for elastic load balancing
The elasticloadbalancing:ResourceTag/key condition key has been specific to elastic
load balancing. Following are the actions supporting this condition key:
API version 2015-12-01
AddTags
• CreateListener
• CreateLoadBalancer
• DeleteLoadBalancer
• DeleteTargetGroup
• DeregisterTargets
• ModifyLoadBalancerAttributes
• ModifyTargetGroup
• ModifyTargetGroupAttributes
• RegisterTargets
• RemoveTags
• SetIpAddressType
• SetSecurityGroups
• SetSubnets
API version 2012-06-01

37A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
AddTags
• ApplySecurityGroupsToLoadBalancer
• AttachLoadBalancersToSubnets
• ConfigureHealthCheck
• CreateAppCookieStickinessPolicy
• CreateLBCookieStickinessPolicy
• CreateLoadBalancer
• CreateLoadBalancerListeners
• CreateLoadBalancerPolicy
• DeleteLoadBalancer
• DeleteLoadBalancerListeners
• DeleteLoadBalancerPolicy
• DeregisterInstancesFromLoadBalancer
• DetachLoadBalancersFromSubnets
• DisableAvailabilityZonesForLoadBalancer
• EnableAvailabilityZonesForLoadBalancer
• ModifyLoadBalancerAttributes
• RegisterInstancesWithLoadBalancer
• RemoveTags
• SetLoadBalancerListenerSSLCertificate
• SetLoadBalancerPoliciesForBackendServer
• SetLoadBalancerPoliciesOfListener
Flexible Load Balancing is an Amazon web administration that improves the
accessibility and adaptability of your application. With Elastic Load Balancing, you can
disperse application stacks between at least two Amazon EC2 occasions. Versatile Load
Balancing improves accessibility through repetition, and it underpins traffic development for
your application.

38A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Flexible Load Balancing consequently disperses and balances approaching application
traffic among all the EC2 server occurrences you are running. The administration additionally
makes it simple to include new cases when you have to build the limit of your application.
Versatile Beanstalk consequently arrangements Elastic Load Balancing when you
convey an application. Under Load Balancing, on the Configuration tab for your condition
inside the Toolkit for Eclipse, you can alter the Elastic Beanstalk condition's heap adjusting
setup.
As a matter of course, a heap balancer courses each solicitation freely to the server
occasion with the littlest burden. By correlation, a sticky session ties a client's session to a
particular server occurrence with the goal that all solicitations originating from the client
amid the session are sent to a similar server occasion.
Versatile Beanstalk utilizes load balancer– produced HTTP treats when sticky
sessions are empowered for an application. The heap balancer utilizes a unique burden
balancer– produced treat to follow the application occasion for each solicitation. At the point
when the heap balancer gets a solicitation, it first verifies whether this treat is available in the
solicitation. Assuming this is the case, the solicitation is sent to the application occurrence
determined in the treat. On the off chance that it finds no treat, the heap balancer picks an
application occurrence dependent on the current burden adjusting calculation. A treat is
embedded into the reaction for restricting consequent solicitations from a similar client to that
application case. The arrangement setup characterizes a treat expiry, which sets up the length
of legitimacy for every treat. Under Load Balancer in the Sessions area, indicate whether the
heap balancer for your application permits session stickiness and the term for every treat.
In light of the information and discoveries assembled from the writing survey
segments have been centered on the discovering the consequences of the targets started. The

39A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
heap balancer is equipped for taking care of the session determination at whatever point
required. The exploration would likewise be related with finding an answer for the holding up
time of the high tedious assignments alongside aiding in diminishing the deferrals in the
activities. The whole examination is related with centering upon the diverse burden offsetting
calculation alongside finding of answers for the various sorts of issues looked by the heap
adjusting. Amazon Web Services versatile burden balancer apparatus have been utilized for
observing burden on the cloud server. A heap balancer helps in circulating approaching rush
hour gridlock crosswise over targets including EC2 cases. This has helped in empowering
increment of accessibility of use. The heap balancer has been utilized in observing the
strength of enlisted targets and guarantees that directing is done to deals in sound targets.
Designing burden balancer for tolerating approaching traffic has been finished with audience
members that are associated with a convention and port number for arranging from customers
to stack balancer. Flexible burden adjusting underpins different sorts of burden balancers
including application load balancers, organize load balancers and great burden balancers. An
application load balancer settles on directing and burden adjusting choices at application
layer (HTTP/HTTPS). A system load balancer helps n directing and load adjusting choices at
the vehicle layer (TCP/TLS). Both application load balancers and system load balancers have
had the option to course demands to at least one ports on each EC2 occurrence in virtual
private cloud (VPC). A Classic Load Balancer settles on directing and burden adjusting
choices either at the vehicle layer (TCP/SSL) or the application layer (HTTP/HTTPS), and
supports either EC2-Classic or a VPC.
A heap balancer helps in tolerating approaching traffic from customers and courses
solicitations to enlisted focuses in at least one Availability Zones. The heap balancer has been
helping in observing soundness of its regeistered targets or guaranteeing directing traffic to
solid targets. On the off chance that, heap balancer recognizes an undesirable target, it quits

40A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
steering traffic to that objective and resume directing traffic to that objective as it identifies
it's as sound.
At the point when an accessibility zone has been empowered for burden balancer,
flexible burden adjusting helps in making load balancer hub in accessibility zone. At the
point when register focuses in an accessibility zone and not ready to empower accessibility
zone, these enlisted targets wear it get traffic. It has been prescribed that various accessibility
zones should be empowered. This assistance on the off chance that one accessibility zone is
inaccessible and has no solid focuses on, the heap balancer still keep on directing traffic to
sound focuses on various accessibility zone. On the off chance that accessibility zone, is
incapacitated, the objectives is that accessibility zone stay enrolled with burden balancer.
The hubs of burden balancer circulate demands from customers for enrolling targets.
At the point when cross zone load adjusting has been empowered, each heap balancer hub
helps in conveying traffic over the enrolled focuses in all empowered accessibility zones. On
the off chance that cross burden adjusting has been empowered, each heap balancer hubs
helps in dispersing traffic over the enrolled focuses in different empowered accessibility
targets. The accompanying graphs show the impact of cross-zone load adjusting. There are
two empowered Availability Zones, with 2 focuses in Availability Zone An and 8 focuses in
Availability Zone B. Customers send solicitations, and Amazon Route 53 reacts to each
demand with the IP address of one of the heap balancer hubs. This conveys traffic to such an
extent that each heap balancer hub gets half of the traffic from the customers. Each heap
balancer hub disperses a lot of the traffic over the enlisted focuses in its extension.

41A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
5. Conclusion
The load balancer is capable of handling the session persistence whenever needed.
The research would also be associated with finding a solution for the waiting time of the high
time-consuming tasks along with helping in decreasing the delays in the operations. The
entire study is associated with focusing upon the different load balancing algorithm along
with finding of solutions for the different types of issues faced by the load balancing. It can
be seen that there are several ways to apply load balancing techniques in cloud computing. To
be more precise, the two most prominent ways are load balancing on virtual servers and load
balancing on physical servers. The paper has specifically focused on load balancing in virtual
servers through cloud computing environments. It typically looks into the policy of allocating
resources from a physical server to different virtual servers or VMs for applications and tasks
that are executing on them. It is more importantly crucial to pay attention to the amount of
load made by the virtual machines. It is essential to be able to execute and run the tasks
properly while at the same time properly concentrating on the load. In other words, the
requirements of the virtual machine should be made accordingly. In a typical hybrid cloud
based environment, the load distribution between the different VMs or virtual machines is
heterogeneous in kind with respect to the processing power. More so, because, it is to be
ensured that each of the virtual machines can have different processing cost and time. An
efficient load balancing algorithm typically focuses on selecting the virtual machine that will
cost the least amount of processing time for assigning tasks.
5.1 Limitation of the Study
The study has focused on load balancing on virtual servers rather than on physical
servers. Moreover, the study only highlights the process of static load balancing through
Amazon EC2. Due to the limited span of time allocated for the study, it was not possible to

42A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
conduct a more thorough and detailed review and critical analysis of the literature. Apart
from that, the paper mainly focuses on the potential benefits of performing load balancing on
cloud computing and does not take into detail consideration of the possible issues or
difficulties that could be faced while executing dynamic load balancing on hybrid cloud
environments.
5.2 Future Scope of the Study
The study can be further extended in future in order to critically analyze the working
principles of the dynamic load balancing algorithms. The paper can also be improved
potentially with the help of practical examples and case studies that have taken place and are
associated with load balancing issues and concerns. In addition to that, other future
possibilities of extending the study also includes identifying and recognizing the best
practices for load balancing in a typical hybrid cloud environment. Apart from that, there are
only few of the load balancing algorithms discussed in this paper. There is a potential for
carrying out future research on the various other available load balancing algorithms such as
CLBDM (Central Load Balancing Decision Model), MapReduce Based Entity Resolution,
INS (Index Name Server Algorithm), OLB (Opportunistic Load Balancing), LBMM (Load
Balancing Min Min) algorithm.

43A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
References
Adhikari, M. and Amgoth, T., 2018. Heuristic-based load-balancing algorithm for IaaS
cloud. Future Generation Computer Systems, 81, pp.156-165.
Aslam, S. and Shah, M.A., 2015, December. Load balancing algorithms in cloud computing:
A survey of modern techniques. In 2015 National Software Engineering Conference
(NSEC) (pp. 30-35). IEEE.
Batrouni, M., Zions, J.D. and Homiou, O., Microsoft Technology Licensing LLC,
2017. CDN load balancing in the cloud. U.S. Patent 9,537,973.
Bhatt, H.H. and Bheda, H.A., 2015, September. Enhance Load Balancing using Flexible
Load Sharing in Cloud Computing. In 2015 1st International Conference on Next Generation
Computing Technologies (NGCT) (pp. 72-76). IEEE.
Chen, S.L., Chen, Y.Y. and Kuo, S.H., 2017. CLB: A novel load balancing architecture and
algorithm for cloud services. Computers & Electrical Engineering, 58, pp.154-160.
Chien, N.K., Son, N.H. and Loc, H.D., 2016, January. Load balancing algorithm based on
estimating finish time of services in cloud computing. In 2016 18th International Conference
on Advanced Communication Technology (ICACT) (pp. 228-233). IEEE.
Dave, A., Patel, B. and Bhatt, G., 2016, October. Load balancing in cloud computing using
optimization techniques: A study. In 2016 International Conference on Communication and
Electronics Systems (ICCES) (pp. 1-6). IEEE.
Devi, D.C. and Uthariaraj, V.R., 2016. Load balancing in cloud computing environment
using improved weighted round robin algorithm for nonpreemptive dependent tasks. The
scientific world journal, 2016.

44A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Farrag, A.A.S., Mahmoud, S.A. and El Sayed, M., 2015, December. Intelligent cloud
algorithms for load balancing problems: A survey. In 2015 IEEE Seventh International
Conference on Intelligent Computing and Information Systems (ICICIS) (pp. 210-216). IEEE.
Gandhi, R., Liu, H.H., Hu, Y.C., Lu, G., Padhye, J., Yuan, L. and Zhang, M., 2015. Duet:
Cloud scale load balancing with hardware and software. ACM SIGCOMM Computer
Communication Review, 44(4), pp.27-38.
Ghomi, E.J., Rahmani, A.M. and Qader, N.N., 2017. Load-balancing algorithms in cloud
computing: a survey. Journal of Network and Computer Applications, 88, pp.50-71.
Gopinath, P.G. and Vasudevan, S.K., 2015. An in-depth analysis and study of Load balancing
techniques in the cloud computing environment. Procedia Computer Science, 50, pp.427-
432.
Kapoor, S. and Dabas, C., 2015, August. Cluster based load balancing in cloud computing.
In 2015 Eighth International Conference on Contemporary Computing (IC3) (pp. 76-81).
IEEE.
Kumar, A. and Kalra, M., 2016, April. Load balancing in cloud data center using modified
active monitoring load balancer. In 2016 International Conference on Advances in
Computing, Communication, & Automation (ICACCA)(Spring) (pp. 1-5). IEEE.
Kumar, P. and Kumar, R., 2019. Issues and Challenges of Load Balancing Techniques in
Cloud Computing: A Survey. ACM Computing Surveys (CSUR), 51(6), p.120.
Mesbahi, M. and Rahmani, A.M., 2016. Load balancing in cloud computing: a state of the art
survey. Int. J. Mod. Educ. Comput. Sci, 8(3), p.64.

45A STUDY OF LOAD BALANCING IN CLOUD COMPUTING
Milani, A.S. and Navimipour, N.J., 2016. Load balancing mechanisms and techniques in the
cloud environments: Systematic literature review and future trends. Journal of Network and
Computer Applications, 71, pp.86-98.
Mousavi, S., Mosavi, A. and Varkonyi-Koczy, A.R., 2017, September. A load balancing
algorithm for resource allocation in cloud computing. In International Conference on Global
Research and Education (pp. 289-296). Springer, Cham.
Oueis, J., Strinati, E.C. and Barbarossa, S., 2015, May. The fog balancing: Load distribution
for small cell cloud computing. In 2015 IEEE 81st Vehicular Technology Conference (VTC
Spring) (pp. 1-6). IEEE.
Paya, A. and Marinescu, D.C., 2015. Energy-aware load balancing and application scaling for
the cloud ecosystem. IEEE transactions on cloud computing, 5(1), pp.15-27.
Singh, A., Juneja, D. and Malhotra, M., 2015. Autonomous agent based load balancing
algorithm in cloud computing. Procedia Computer Science, 45, pp.832-841.
Venkiteswaran, M. and Shah, M.D., A10 Networks Inc, 2018. Load Balancing Between
Computing Clouds. U.S. Patent Application 15/476,572.
Xu, M., Tian, W. and Buyya, R., 2017. A survey on load balancing algorithms for virtual
machines placement in cloud computing. Concurrency and Computation: Practice and
Experience, 29(12), p.e4123.
Zhao, J., Yang, K., Wei, X., Ding, Y., Hu, L. and Xu, G., 2015. A heuristic clustering-based
task deployment approach for load balancing using Bayes theorem in cloud
environment. IEEE Transactions on Parallel and Distributed Systems, 27(2), pp.305-316.