Load Balancing in Cloud Computing

Laraib Gul,

BSCSF18M002,
Department of CS and IT,
University of Sargodha,
Sargodha.

Abstract—The past few years have witnessed the emergence of algorithm will make sure that every node in the system does
a novel paradigm called cloud computing. Cloud computing is more or less same volume of work. Load balancing is a new
a structured model that defines computing services, in which approach that assists networks and resources by providing a
data as well as resources are retrieved from cloud service high throughput and least response time. In cloud platforms,
provider via internet through some well-formed web-based tool
resource allocation (or load balancing) takes place majority
and application. Cloud Computing is nothing but a collection
of computing resources and services pooled together and is at two levels.
provided to the users on pay-as-needed basis. Sharing of the  At first level: The load balancer assigns the
group of resources may initiate a problem of availability of requested instances to physical computers at the
these resources causing a situation of deadlock. One way to time of uploading an application attempting to
avoid deadlocks is to distribute the workload of all the VMs balance the computational load of multiple
among themselves. This is called load balancing. Load applications across physical computers.
unbalancing problem is a multi-variant, multi-constraint
problem that degrades performance and efficiency of  At second level: When an application receives
computing resources. Load balancing algorithms and job multiple incoming requests, each of these requests
allocations are main research problems in areas of resource must be assigned to a specific application instance
management of future internet. The goal of balancing the load to balance the computational load across a set of
of virtual machines is to reduce energy consumption and instances of the same application.
provide maximum resource utilization thereby reducing the This paper is organized as follows: Section 2 contains an
number of job rejections. The aim of this paper is to discuss the introduction of load balancing. Section 3 describes the
concept of load balancing in cloud computing. This paper deals existing classification of load balancing algorithms. Dynamic
with the different load balancing algorithms such as static load load balancing policies are analyzed in section 4. Different
balancing algorithms (SLBA) and dynamic load balancing
load balancing metrics are given in section 5 and final
algorithms (DLBA). The load balancing algorithm can be used
conclusion of the work in section 6.
to better utilize and better understand user needs.
II. LOAD BALANCING
Keywords— Cloud Computing, Load balancing, load balancer,
static load balancing, dynamic load balancing algorithm, load 2.1 Overview of Load Balancing:
balancing metrics.
Load balancing is the process of improving the performance
I. INTRODUCTION of the system by shifting of workload among the processors.
Workload of a machine means the total processing time it
The US National Institute of Standards and Technology requires to execute all the tasks assigned to the machine.
(NIST) characterizes cloud computing as pay-per-use model Balancing the load of virtual machines uniformly means that
for enabling available, convenient, on-demand network anyone of the available machine is not idle or partially
access to a shared pool of configurable computing resources loaded while others are heavily loaded. In cloud computing,
(e.g. networks, servers, storage, applications, services) that if users are increasing load will also be increased, the
can be rapidly provisioned and released with minimal increase in the number of users will lead to poor performance
management effort or service provider interaction. A cloud in terms of resource usage. With the load balancer you can
computing model is efficient if its resources are utilized in split the workload and balance it between two or more
best possible way and such an efficient utilization can be servers in the cloud. As a result, you can configure your
achieved by employing and maintaining proper management infrastructure to maximize activity, optimize resource
of cloud resources. The problem of load unbalancing is an allocation, and provide a smallest amount of response time.
undesirable event in the Cloud service provider side that If some good load balancing technique is implemented, it
degrades the performance and efficacy of the computing will equally divide the load (here term equally defines low
resources along with guaranteed Quality of Service (QoS) on load on heavy loaded node and more load on node with less
agreed Service Level Agreement (SLA) between consumer load now) and thereby we can maximize resource
and provider. Under these circumstances there arises need for
load balancing (LB) and is a peculiar topic of research
interest among researchers. The load balancing in cloud
computing can be done at physical machine level or VM
level. Load balancing is a method that distributes the
workload among diverse nodes in the given environment
such that it ensures no node in the system is over loaded or
sits idle for any instant of time. An efficient load balancing
utilization. directly proportional to each other. Load balancing helps in
reducing energy consumption which will automatically
reduce carbon emission and thus achieve Green Computing.

2.4 Significance of Load Balancing:

i. Better Performance:
Load balancing methods are smaller and easier to implement
than their counterparts. Organizations can work their
Fig 1. Diagram for load balancing customers' applications more quickly and deliver better
performance at relatively low cost.
ii. Maintain Website Traffic:
2.2 What is meant by load balancer? Cloud balancing provides scalability to manage your site
The main aim of the load balancer helps to assign resources traffic. With efficient load balancing, you can easily manage
equally to the tasks for resource efficiency and user user traffic at a high level with the presence of servers and
satisfaction at minimal expense, quality output, gripping network devices.
rapid traffic blast sustain traffic on the website and elasticity iii. Handle Sudden Traffic Burst:
which motivates us to identify problems in LB and to work Load balancers have this ability to handle any sudden traffic
on their resolution. received at a specified time. Load balancers distribute the
entire site load on different servers to get optimal results
with minimal response time.
iv. Flexibility:
If the workload is distributed across servers or network
devices, even if one node fails, the load can be replaced by
another node. It shows scalability, flexibility, and the ability
to handle traffic.

2.5 Activities involved in load balancing:

Fig 2. Illustration of cloud load balancer

Using this type of configuration, the load balancer accepts

network requests and sorts them between two or more
servers in the cloud.

2.3 Goals of Load Balancing:

Goals of load balancing include:
 Stability maintenance of the system,
 Increase flexibility of the system so as to adapt to
the modifications,
 Build a fault tolerant system by creating backups. Fig 3. Two level Load Balancing Architecture
 Reduce waiting time,
 Reduce the response time, The request generator generates user requests which are
user tasks that need computing resources for their execution.
 Increase the utilization of resources,
Data center controller is in-charge of task management.
 Improve reliability , The load balancer checks which VM to assign for a given
 Increase throughput, user task. The first level load balancer balances the given
 Load balancing enhance the performance of the workload on individual Physical Machines by distributing
system by managing every node. the workload among its respective associated Virtual
Machines. The second level load balancer balances the
Load balancing is also needed for achieving Green workload across different Virtual Machines of different
computing in clouds. The factors responsible for it are: Physical Machines.
Scheduling and allocating tasks to VMs based on their
1. Limited Energy Consumption: Load balancing can requirements constitute the cloud computing workload. The
reduce the amount of energy consumption by avoiding over load balancing process involves the following activities:
hearting of nodes or virtual machines due to excessive
workload. 2.5.1. Identification of user task requirements:
2. Reducing Carbon Emission: Energy consumption and This phase identifies the resource requirement of the user
carbon emission are the two sides of the same coin. Both are tasks to be scheduled for execution on a VM.
2.5.2. Identification of resource details of a VM: Based on the current state of the system load balancing
This checks the status of resource details of a VM. It gives algorithms are classified as:
the current resource utilization of VM and the unallocated 1. Static Load Balancing
resources. Based on this phase, the status of VM can be
determined as balanced, overloaded or under-loaded with In the static load balancing algorithm the decision of
respect to a threshold. shifting the load does not depend on the current state of the
2.5.3. Task scheduling: system. It requires knowledge about the applications and
Once resource details of a VM are identified the tasks are resources of the system. The performance of the virtual
scheduled to appropriate resources on appropriate VMs by a machines is determined at the time of job arrival. The master
scheduling algorithm. processor assigns the workload to other slave processors
according to their performance. The assigned work is thus
2.5.4. Resource allocation:
performed by the slave processors and the result is returned
The resources are allocated to scheduled tasks for execution.
to the master processor.
A resource allocation policy is being employed to
accomplish this. While, scheduling is required for speeding Static load balancing algorithms are not preemptive and
up the execution, allocation policy is used for proper therefore each machine has at least one task assigned for
resource management and improving resource performance. itself. This algorithm has a drawback that the task is assigned
The strength of the load balancing algorithm is determined to the processors or machines only after it is created and that
by the efficacy of the scheduling algorithm and the task cannot be shifted during its execution to any other
allocation policy. machine for balancing the load.
2.5.5. Migration: 2. Dynamic Load Balancing
Migration is an important phase in load balancing process in In this type of load balancing algorithms the current state of
cloud and latter is incomplete without the former. Migration the system is used to make any decision for load balancing,
is of two kinds in cloud based on entity taken into thus the shifting of the load depends on the current state of
consideration- VM migration and task migration. VM the system. It allows for processes to move from an over
migration is the movement of a VM from one physical host utilized machine to an underutilized machine dynamically
to another to get rid of the overloading problem and is for faster execution.
categorized into types as live VM migration and non-live This means that it allows for process preemption which is
migration. Likewise task migration is the movement of not supported in Static load balancing approach. An
tasks across VMs and is of two types: intra VM task important advantage of this approach is that its decision for
migration and inter VM task migration. An efficient balancing the load is based on the current state of the system
migration technique leads to an efficient load balancing. which helps in improving the overall performance of the
From the extensive survey it has been concluded that task system by migrating the load dynamically. Traditional types
migration process is more time and cost effective than VM of algorithms come under the static ones and the meta-
migration and the trend has shifted from VM to task heuristic algorithms come under the dynamic algorithms.
migration.
3.1 Static Load Balancing Algorithms:
Round-robin algorithms: presently it is available such as (1)
III. CLASSIFICATION OF LOAD BALANCING ALGORITHMS round robin, (2) weighted round robin.
In cloud computing, cloud servers should always be
balanced, to use the resources with their full capacity. 3.1.1 Round Robin:
Sometimes it happens that some servers are heavily loaded Round robin use the time slicing mechanism. The name of
while the other servers are under loaded or in idle state. To the algorithm suggests that it works in the round manner
overcome this problem load balancing algorithms are used. where each node is allotted with a time slice and has to wait
These algorithms help in allocating every single task by for their turn. The time is divided and interval is allotted to
monitoring load on each server. According to the Deepak each node. Each node is allotted with a time slice in which
Mahapatra, the balancing algorithm is defined as “The load they have to perform their task. The complicity of this
balancing in clouds may be among physical hosts or VMs. algorithm is less compared to the other two algorithms. An
This balancing mechanism distributes the dynamic workload open source simulation performed the algorithm software
evenly among all the nodes (hosts or VMs). The load
know as cloud analyst, this algorithm is the default
balancing in the cloud is also referred to as load balancing as
algorithm used in the simulation. This algorithm simply
a service (LBaaS)”.
allots the job in round robin fashion which doesn't consider
Based on process orientation load balancing algorithms the load on different machines.
are classified as:
a) Sender Initiated: In this sender initiates the process; the
client sends request until a receiver is assigned to him to
receive his workload.
b) Receiver Initiated: The receiver initiates the process;
the receiver sends a request to acknowledge a sender who is
ready to share the workload.
c) Symmetric: It is a combination of both sender and
receiver initiated type of load balancing algorithm.
 (hence min-min) to a node that is capable of handling it.
Architectural description of Min-Min algorithm is shown
below:

Fig 4. Round Robin algorithm

3.1.2 Weighted Round Robin:

It was developed to overcome the drawback of round robin

algorithm. Here it assigns weight to the servers according to
their load handling capacity with respect to the weight tasks
are allocated. This algorithm helps to maintain the load on
the server as well as their capacity. If two tasks are assigned
to two different servers then the third task will be allocated
to the higher weightage server.

Fig.6 Architectural description of min-min algorithm.

3.1.4 Max-Min load balancing algorithm:

Fig 5. Weighted Round Robin algorithm The max-min algorithm is much the same as to min-min
algorithm. At first for all the available tasks are submitted to
Time-based algorithms: are like (1) Max–Min algorithm, (2) the system and minimum completion time for all of them are
Min–Min algorithm; calculated, then among these tasks the one which is having
the completion time, maximum is chosen and that is
3.1.3 Min-Min load balancing algorithm: allocated to the corresponding machine. This algorithm
The Algorithm take up with a task set which are initially outperform than Min-Min algorithm where when short tasks
not assigned to any of the nodes. Initially the minimum are in high numbers when compared to that of long ones.
completion time is calculated for all the available nodes. For e.g. if in a task set only a single long task is presented
Once this calculation gets completed the task having the then ,Max Min algorithm runs short tasks concurrently
completion time minimum is chosen and assigned to the along with long task. The make span focus on how much
respective node. The execution time of all other tasks which small tasks will get executed concurrently with the large
are currently available in that machine is updated and the ones. Max-Min is almost identical to Min-Min, except it
task gets discarded from the available task set. The routine selects the task having the maximum completion time and
is done time after time until all the tasks have been assigned allocates to the corresponding machine. The algorithm
to the equivalent machines. The algorithm works better suffers from starvation where the tasks having the maximum
when the situation is like where the small tasks are greater completion time will get executed first while leaving behind
in number of than the large tasks. The algorithm has a the tasks having the minimum completion time.
disadvantage that it leads to starvation. Architectural description of Max Min algorithm is presented
Min-Min is a simple and fast algorithm capable of providing below:
improved performance. Min-Min schedules the ideal tasks at
first which results in best schedules and improve the overall
make span. Assigning small task first is its drawback. Thus,
smaller tasks will get executed first, while the larger tasks
keeps on in the waiting stage, which will finally results in
poor machine use. Min-Min exhibits minimum completion
time for jobs which are unassigned (similar to MCT), and
later allocating the jobs with minimum completion time
 3.1.7 Weighted Least Connection:
This algorithm is similar to the least connection algorithm.
As seen in Weighted Round Robin where weights are
assigned to the server likewise in Weighted Least
Connection also weights are assigned in numerical form.

3.2 Dynamic Load Balancing Algorithms:

Optimization algorithms: are available such as (1) honey bee
foraging, (2) ant colony optimization, and (3) hybrid genetic-
based host load and also in particle swarm optimization.

3.2.1 Honey bee foraging:

This dynamic algorithm is derived from the honey bees; the
detailed study is done on their behavior of how they search
for food and reap food. Scout is a class of bees which finds
food and informs others through the dance called vibration.
This gives the estimate about the quality and quantity of the
food. This technique is used in load balancing to inform the
underloaded and overloaded VMs. The tasks from the
overloaded machines are shifted to the underloaded
machine. Similarly, the tasks from the overloaded VMs are
considered as the honey bees. Submitting tasks to the
underloaded VM, the task will update the number of tasks
and load of that particular VM to all other waiting tasks.
Based on the load and the priority tasks choose VMs.
Whenever a high priority task has to be submitted to other
VMs, it should consider the VM that has a minimum
Fig .7 Architectural description of max-min algorithm. number of high priority tasks so that the particular task will
be executed at the earliest. Since all VMs will be sorted in
ascending order based on load, the task removed will be
Agent-based algorithm includes algorithms such as
submitted to under loaded VM. In essence, the tasks are the
Opportunistic Load Balancing. honey bees and the VMs are the food sources. Loading of a
task to a VM is similar to a honey bee foraging a food
3.1.5 Opportunistic Load Balancing (OLB): source.

It keeps all the servers busy and never considers the load of 3.2.2 Ant colony optimization:
the task which is currently running on the server. Besides ACO is a technique of problem-solving inspired by the
the current task running on the server it randomly allocated behavior of ants in searching the optimal paths from the nest
another task. to food; they all work together and search new sources of
food while some ants parallel works on shifting food from
Connection-based algorithms: are available such as (1) least source to nest. Many researchers are inspired by this
connection, (2) weighted least connection. behavior of ants and helped them to solve real life problems
in different fields. In this method, a pheromone table of a
3.1.6 Least Connection: node is maintained and an ant updates the entries of the
In this algorithm it considers currently running tasks also node from source to destination. The other routing ants
which is not taken by the above mentioned algorithms. reference the pheromone table and calls that have it as their
According to which other tasks are assigned, these lead to destination. However, for asymmetric networks, the costs
the least number of active sessions in current time. from to and from to may be different. Hence, in this
approach for updating pheromone is only appropriate for
routing in symmetric networks.

3.2.3 Genetic Algorithms (GA):

GA is one of the most used algorithms which solves the NP
(Non-polynomial) -complete problems. It is derived from
the soft computing method. It comes under the heuristic
search process. GA is inspired by natural evolution from the
human mind and genetics.
A simple GA consisting of triple processes: (1) availability,
Fig 8. Least Connection Algorithm (2) Genetic and (3) replacement operations. GA's core is the
creation of offspring via mutations & crossover methods,
with the assumption that either binary coding, tree coding or allocate the process to the lightly loaded instance of that
numerical coding depends on the type of the chromosome. group.

3.2.4 Heuristic Algorithm: IV. DYNAMIC LOAD BALANCING POLICIES

Traditional methods of solving problems were very slow The different policies are described as follows:
and not compatible for solving NP-complete problems.
1. Location Policy: The policy used by a processor or
Heuristic algorithms solve problems in a faster and efficient
machine for sharing the task transferred by an over loaded
way than the traditional way and designed to solve decision machine is termed as Location policy.
problems. Example is the Traveling Salesmen Problem
where there is a list of cities to visit and the distance 2. Transfer Policy: The policy used for selecting a task or
between the two cities are given and the optimal solution is process from a local machine for transfer to a remote machine
to visit all the cities by travelling a minimum distance. is termed as Transfer policy.
3. Selection Policy: The policy used for identifying the
3.2.5 Particle Swarm Optimization (PSO) Algorithm: processors or machines that take part in load balancing is
PSO was first introduced by Kennedy and Eberhart; it is termed as Selection Policy.
one type of a meta-heuristics method. The PSO algorithm is
similar to other population-based algorithms like GA but, 4. Information Policy: The policy that is accountable for
gathering all the information on which the decision of load
there is no direct recombination of individuals of the
balancing is based is referred as Information policy.
population. PSO follows a model or a pattern for the social
interaction and creates the communication between them, 5. Load estimation Policy: The policy which is used for
for example, fish schools and bird flock. PSO algorithm is deciding the method for approximating the total work load of
very similar to the swarm intelligence optimization a processor or machine is termed as Load estimation policy.
algorithms. It mainly focuses on reducing the total cost of 6. Process Transfer Policy: The policy which is used for
computation of an application on the cloud computing deciding the execution of a task that is it is to be done locally
environment. or remotely is termed as Process Transfer policy.
7. Priority Assignment Policy: The policy that is used to
3.2.6 Hill climbing:
assign priority for execution of both local and remote
A Hill climbing is the simple and iterative method based processes and tasks is termed as Priority Assignment Policy.
optimization algorithm; it moves towards the best solution
step by step. It selects the increasing value and moves 8. Migration Limiting Policy: The policy that is used to set
towards the peak or uphill. Peak is the top of the hill or the a limit on the maximum number of times a task can migrate
place where no neighboring value is more than the peak. from one machine to another machine.
The algorithm stops once it reaches the peak or to the
V. LOAD BALANCING METRICS
stopping criteria. Sometimes it can reach a local optimum
solution instead of global optimum. For the load balancing
this algorithm maintains an index table with the list of VMs After studying the dynamic load balancing algorithms, we
and their states (BUSY/AVAILABLE). When the new task have compared all the algorithms on the bases of some
request arrives randomly generates VM id and allocates the predefined metrics. These metrics are as follows:
task to the VM if the state is available otherwise generates Throughput: Throughput is used to calculate the number of
another VM id randomly. It updates the index table jobs whose execution has been completed. It should be high
accordingly. Once the task is completed, de-allocates the to improve the performance of the system.
VM and updates the table.
Overhead: It determines the amount of overhead involved
while implementing a load balancing algorithm. Overhead
3.2.7 Equally Spread Current Execution (ESCE): should be minimized so that a load balancing technique can
ESCE is a dynamic load balancing algorithm, which work efficiently.
handles the process with priority. It determines the priority
by checking the size of the process. This algorithm Fault Tolerance: Fault tolerance system is a system in which
distributes the load randomly by first checking the size of the processing does not get affected because of the failure of
the process and then transferring the load to a virtual any particular processing device in the system. The load
machine, which is lightly loaded. The load balancer spreads balancing should be fault tolerant.
the load on different nodes, and hence, it is known as spread Migration time: Migration is the time of movement of job of
spectrum technique. the master system to the slave system and vice versa in case
of results. Migration time is the overhead, which cannot be
3.2.8 Throttled Load Balancer (TLB): removed but should be minimized.
Throttled load balancer is a dynamic load balancing
Response Time: It is the amount of time taken to respond by
algorithm in which the client first requests the load balancer
a particular load balancing algorithm in a distributed system.
to find a suitable virtual machine to perform the required This parameter should be minimized.
operation. In Cloud computing, there may be multiple
instances of virtual machine. These virtual machines can be Resource Utilization: It is used to check the utilization of
grouped based on the type of requests they can handle. resources. It should be optimized for an efficient load
Whenever a client sends a request, the load balancer will balancing.
first look for that group, which can handle this request and
Scalability: It is the ability of an algorithm to perform load [7] Parin. V. Patel, Hitesh. D. Patel, Pinal. J. Patel, “A Survey on Load
balancing for a system with any finite number of nodes. This Balancing in Cloud Computing” IJERT, Vol. 1, Issue 9, November
2012.
metric should be improved.
[8] R. Mata-Toledo, and P. Gupta, “Green data center: how green can we
Performance: It is used to check the efficiency of the system. perform”, Journal of Technology Research,Academic and Business
This has to be improved at a reasonable cost, e.g., reduce task Research Institute, Vol. 2, No. 1,May 2010, pages 1-8.
response time while keeping acceptable delays. [9] S. K. Garg, C. S. Yeob, A. Anandasivamc, and R.
Buyya,“Environment-conscious scheduling of HPC applications
ondistributed Cloud-oriented data centers”, Journal of Parallel and
VI. CONCLUSION Distributed Computing, Elsevier, Vol. 70, No. 6, May 2010, pages 1-
The cloud computing has the dynamic nature and due to 18.
which cloud network has various issues like security, quality [10] Jitendra Bhatia, Tirth Patel, Harshal Trivedi, Vishrut Majmudar,” HTV
Dynamic Load Balancing Algorithm for Virtual Machine Instances in
of service and fault occurrence etc. The load balancing is the Cloud”,18,Dec2 012,Pages 15-20 IEEE.
major issue of cloud network which reduce its efficiency. [11] Dr. Hemant S. Mahalle, Prof. Parag R. Kaveri ,Dr.Vinay Chavan,”
Load balancing is a main task in cloud computing for Load Balancing On Cloud Data Centres” , International Journal of
efficient utilization of resources. The main goal of load Advanced Research in Computer Science and Software
balancing is to achieve higher client satisfaction, Engineering,Vol.3,Issue 1,January 2013.
maximize resource utilization and increase the performance [12] Cisco (2009) Cisco visual networking index: Forecast and
methodology, 2009-2014.White paper.
of the cloud system thereby reduction in the energy
[13] Weiss A (2007) Computing in the clouds. Networker 11: 16-25.
consumption and carbon emission rate. This research
[14] Hayes B (2008) Cloud computing. Commun ACM 51: 9-11.
described the numerous algorithms for LB such as static load
[15] Mell P, Grance T (2009) Draft NIST Working Definition of Cloud
balancing algorithms and dynamic load balancing Computing. Nat Inst Standards Technol
algorithms. Although different dynamic and static algorithms [16] Rimal, Prasad B, Choi E, Lumb V (2009) A taxonomy and survey of
are proposed and developed in recent time each one has own cloud computing systems. Proceedings of 5th International Joint
advantages and disadvantages. In this manner, the objective Conference on INC, IMS and IDC, IEEE .
is to balance the audience traffic of the cloud infrastructure [17] Sinha PK (1997) Distributed operating Systems Concepts andDesign.
while enhancing the execution, reducing the overall response IEEE Computer Society Press.
time, increasing the throughput for a particular number of [18] https://kemptechnologies.com/in/load-balancer/load-balancing-
algorithms-techniques/
jobs, and proficiently handle resource usage.
[19] Wang SC, Chen CW, Yan KQ, Wang SS (2013) The Anatomy Study
of Load Balancing in Cloud Computing Environment. The Eighth
International Conference on Internet and Web Applications and
VII. REFERENCES Services 230-235
[1] JianzheTai,JueminZhang,JunLi,WaleedMeleis and NingfangMi “A R [20] Gulati A, Chopra RK (2013) Dynamic Round Robin for Load
A: Adaptive Resource Allocation for Cloud Computing Environments Balancing in a Cloud Computing, International Journal of Computer
under Bursty Workloads” 978-1-4673-0012-4/11 ©2011 IEEE. Science and Mobile Computing 2: 274-278.
[2] Ali M Alakeel, “A Guide To Dynamic Load Balancing In Distributed [21] Khan Z, Singh R, Alam J, Saxena S (2011) Classification of Load
Computer Systems”, International Journal of Computer Science and Balancing Conditions for parallel and distributed systems.
Network Security, Vol. 10 No. 6, June 2010. International Journal of Computer Science Issue 8: 411- 419.
[3] Abhijit A Rajguru, S.S. Apte, “A Comparative Performance Analysis [22] Ragmania, Awatif, et al. "An improved Hybrid Fuzzy-Ant Colony
of Load Balancing Algorithms In Distributed Systems Using Algorithm Applied to Load Balancing in Cloud Computing
Qualitative Parameters”, International Journal of Recent Technology Environment." Procedia Computer Science 151 (2019): 519-526.
and Engineering, Vol. 1, Issue 3, August 2012. [23] Mishra, Sambit Kumar, Bibhudatta Sahoo, and Priti Paramita Parida.
[4] Nidhi Jain Kansal, Inderveer Chana, “Cloud Load Balancing "Load balancing in cloud computing: a big picture." Journal of King
Techniques: A Step Towards Green Computing”, IJCSI, Vol. 9, Issue Saud University-Computer and Information Sciences (2018).
1, January 2012. [24] LD, Dhinesh Babu, and P. Venkata Krishna. "Honey bee behavior
[5] R. Shimonski, Windows 2000 And Windows Server 2003, Clustering inspired load balancing of tasks in cloud computing environments."
and Load Balancing Emeryville, McGrow-Hill Professional Applied Soft Computing 13.5 (2013): 2292-2303.
Publishing, CA, USA, 2003.
[6] David Escalnte and Andrew J. Korty, “Cloud Services: Policy and
Assessment”, EDUCAUSE Review, Vol. 46, July/August 2011.