LOAD BALANCING IN CLOUD COMPUTING

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 04 | Apr 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 333
LOAD BALANCING IN CLOUD COMPUTING
Ayush Agrawal1, Devesh Katiyar2, Gaurav Goel3
1Ayush Agrawal Department of Computer Science, Dr. Shakuntala Misra National Rehabilitation University,
Lucknow, India
2,3Assistant Professor, Faculty of Computer & Information Technology, Dr. Shakuntala Misra National
Rehabilitation University, Lucknow, India
-----------------------------------------------------------------------***--------------------------------------------------------------------------
Abstract - Cloud computing offers us the way to share
data and provides many resources to users. There is an
important issue in cloud computing which is Load
Balancing. With remarkable gain in users and their need
of requests on the cloud computing platform, with lots of
usage of resources became a severe concern. Load
balancing gives user gratification and resource
utilization ratio by guaranteeing an efficient and
impartial allocation of all type of resources. Load
balancing method is used to distribute tasks from high
loaded resources to low loaded or with in the constant
resources There are many algorithms which gives the
user more satisfying experience on the cloud services. In
this paper, we used different type of algorithms to solve
the issues in load balancing.
Keywords: Cloud Computing, Load balancing, Load
balancing Algorithms.
1. Introduction
Cloud Computing has become the important
requirement for the IT companies. It has moved
hardware resources and software services that are on
the internet rather than the resources which are present
at customer end. The user only has to pay for that service
only which they use on cloud. Because of its satisfying
services many organizations are seeing forward to use it.
Due to the high demand of services provided on cloud
allows the user to use the technology, continuing
deployment, and the development of the many
organizations. The Services in Cloud allows many
organizations to increase their investments that is
related with costly data storages and applications and
minimizing these expenses which are required to use
the services on cloud. Today there are many cloud
services providers which are available like- Cloud Stack,
OpenStack, EMC2, AWS Amazon, Google Cloud, Open
Nebula etc.
I. Cloud computing has different types of
features:
 On request service- The user can request for
any services on cloud and can assess anytime.
 Broad Network Access- There are many
services present on the cloud which can access
over internet. The cloud services are generally
get through local networks or on any standard
devices.
 Fast Elasticity- The services on this allows less
usage of running of workloads which required
large number of servers but only for a less
period of time.
 Resource Sharing- There are different models
by which users can share their resources
which are provided by the service provider. All
the resources are dynamically allocated and
reallocated based on the user’s request.
II. There are different types of challenges in
Cloud Services:
1. Interoperability
2. Lack of Experience
3. Performance Monitoring
4. Cost Management
5. High Dependence On Internet
Load balancing is one of the important issue in cloud
computing. Load balancing involves the process of
distribution of different load at various nodes to improve
job response time and system utilization. It also helps in
the situation when any node is overloaded while other
are less loaded or become idle. Load balancing makes
sure that all the systems or nodes performs uniformly
and complete similar amount of work simultaneously.
Since the demand and users for cloud services escalates,
the need of load balancing rises proportionally. It helps
the users to incur high resource utilization and user
satisfaction. Load Balancing is responsible to map all the
work set for the cloud to free the resources and make
them available to enhance the response time and provide
better utilization of the resources. Multiple servers or
multiple resources capable of fulfilling user demands are
required to reach load balancing. Even if one or more
component fails to provide service, load balancing

enables to keep providing serviced by splitting the load
on the other available resources, it helps supplying the
requests of users without fail. It makes sure every
component/resource is lay out equally to provide better
responses. It reduces response time, provides scalability
and restricts blockages. The figure below explains the
development of load balancing in Cloud Computing.
Figure 1 Load Balancing in Cloud Computing
2. Load balancing classification:
Load balancing is basically classified into two categories:
static and dynamic load balancing:
1) Static algorithm: - This approach is mostly
defined in the design or implementation of the
system. Static load balancing algorithms split the
traffic equally between all the servers.
2) Dynamic algorithm: - This approach measured
only the current state of the system during load
balancing results. A dynamic algorithm
rearranges the processes among processors
during execution time, Dynamic approach is
more appropriate for widely distributed systems
such as cloud computing. Major disadvantage of
Dynamic algorithms is the run-time overhead
due to the transmission of load information
among processor and decision-making for the
variety of processes and communication stays
associated with the task rearrangement itself.
Dynamic load balancing algorithms can be
central or distributed, depending on whether the
accountability for task of global dynamic
scheduling should actually reside in the single
processor (centralized) or the work involved in
making conclusions should be physically
distributed among processors.
Dynamic load balancing algorithm have two kinds.
Which are distributed approach and non-distributed
(centralized) approach. It is defined as following:
a) Centralized approach: - In centralized
approach, only a solo node is responsible for
working and distribution within the entire
system. Further all nodes are not responsible for
this.
b) Distributed approach: - In distributed
approach, each node individualistically builds its
individual load vector. Vector assembling the
load data of other nodes. All conclusions are
made nearby using local load vectors. Distributed
approach is more appropriate for generally
distributed systems such as cloud computing.
3. Main goals of load balancing algorithms
1. Cost effectiveness: Load balancing helps in offer
improved system performance at very lower cost.
2. Scalability and flexibility: The system for which
load balancing algorithms are executed may change
their size after some time. So these type of algorithm
must handle these type’s conditions. So that the
algorithm can be scalable and flexible.
3. Priority: Arrangement of the resources or jobs
needs to be done. So that higher significance jobs get
well chance to execute.
4. Load Balancing Algorithms:
a) Round Robin Algorithm:
Round Robin algorithm uses the method of time slice
mechanism. In the Process of this Type of mechanism
time is spread into several slices and specific node is
given a specific time interval or time quantum and
because of this quantum the node will perform its
processes. The resources for this type of service provider
are delivered to the client on the basis of this time
quantum. This algorithm just assigns the jobs in round
robin technique which doesn’t affect the load on
different machines. As of result, at any moment some
node may have heavy load and other node have no
request.
In Round Robin Algorithm the time quantum has a very
significant role for scheduling, because if the process of
time quantum is very large then Round Robin Scheduling
Algorithm became same as of the FCFS Scheduling. If the
time quantum is very small, then the method of Round
Robin Algorithm is called as Processor Sharing
Algorithm and quantity of context switches becomes
very high.

b) Equally Spread Current Execution Algorithm
(ESCE):
In the Process of spread spectrum method, the load
balancer makes effort to reserve the same load on all
types of virtual machines connected with the data centre.
Load balancer keeps an index table of Virtual machines
along with the number of requests currently allocated to
the Virtual Machine. If any type of request arises from
the data centre to allocate the new VM, it tests the index
table for minimum loaded VM. If their case arises where
more than one VM is found than first known VM is
selected for considered for handling the request of the
client/node and the load balancer also returns the VM Id
to the respective data centre controller. The data centre
transfers the request to the VM recognized by that id.
Now the data centre reviews the index table by
increasing the share count of identified VM. When the
allocated task is accomplished by the VM a request is
shifted towards the data centre which is further reported
by the load balancer. The figure 2 shows the Function of
ESCE algorithm.
Figure 2 ESCE Algorithm
c) Min-Min Algorithm:
These type of algorithm starts with a task set which are
originally not allocated to any of the nodes. At first, the
least completion time is considered for all the available
nodes. Then, the task which has the least expected
completion time is selected and allocated to the node
with minimum execution time. Now from the task set the
task is removed. This process is continual until all types
of the tasks have been allocated to the same nodes.
Hence the algorithm become better if the larger task is
smaller than the small task.
d) Max-Min Algorithm:
This type of max-min algorithm is just like min-min
algorithm. Max-Min algorithm starts with the set of all
the submitted tasks in the task-set which are originally
unassigned to some node. At start, the least completion
time for all types of available tasks is assessed. Then the
process with has the extreme execution time is then used
and allotted to the resource with minimum response
time.
This algorithm overtakes the Min-Min algorithm where
short process is in large in numbers as compared to long
ones.
e) Throttled Algorithm:
This types of algorithm works by discovering the
suitable virtual machine for allocation of a particular
task. These type of algorithm the load balancer keeps an
index table of virtual machines along with their states.
The customer first makes a demand to Data Centre to
find an appropriate virtual machine to perform required
particular operation. The Data Centre obtains the
request from customer for the distribution of Virtual
Machine. Then, Data Centre inquiries the load balancer
for distribution of Virtual Machine. The load balancer
finds the index table from above until the first accessible
VM is found or index table is searched fully.
If VM starts, then the VM Id is forwarded to the Data
Centre. Then the Data Centre connects the demand to the
VM recognised by the Id. Later, the Data Centre
recognizes the load balancer of the latest allocation and
then the data Centre studies the index table thoroughly.
During dealing out with the request of the customer, if
VM is not available, then the load balancer gives value -1
to the data Centre. Then the Data Centre lines the
request until the next accessibility of Virtual Machine.
When the VM completes the processing request, it give
results to the data Centre and recognizes load balancer
for the de-allocation of VM. Then the Load balancer
informs the allocation table by reducing the allocation
for VM by 1.
Figure 3 Throttled Algorithm

f) Ant Colony Optimization Algorithm:
This type of algorithms behaves similar like the
behaviour of real ants. It is basically based on the
capacity of ants to search an ideal path from shell to the
source. In Ant Colony Optimization algorithm, when the
demand is instigated, the ant starts its movement. Ants
instigate from the root node and starts moving from one
node to another node and also check whether the node is
or under loaded or overloaded. When ants travel
towards the network, they inform the pheromone table
which saves the data of each node’s operation.
g) Honeybee Foraging Behaviour:
This type of algorithm is a nature inspired Algorithm for
self-organization. Honeybee attains over-all load
balancing via local server actions. The performance of
the system is improved with the increase in system
range. The main drawback is that throughput will not
increase with the increase in size of the system. When
the various population of service types is required then
this type of algorithm is appropriate.
5. Performance Measurement for Load
Balancing:
1. Throughput: - It is used to analyse all tasks whose
implementation has been completed. The enactment
of any system is enhanced when throughput is max.
2. Fault Tolerance: -It means regaining from failure.
The load balancing should be a better fault tolerant
technique.
3. Response Time: - It’s the amount of time that is
engaged by a definite load balancing algorithm to
response a task within a system. This limitation must
be reduced for better performing of a system.
4.Overhead: It occurs because of more time usage in
traveling from one machine to another. It should be
minimized for efficient working of an algorithm.
5.Resource utilization: It is used to see the utilization
of resources in a system. Resources must be utilized
optimally by a load balancing algorithm.
6. Scalability: - It is the capability of an algorithm to
execute Load balancing for any finite number of
nodes of a system. This metric must be improved for
the system.
7.Performance: It is the efficiency of system during
the load balancing. Performance must be improved
by educing response time, by increasing throughput
and at a reasonable cost.
Table 1- Comparison of different load balancing
algorithms based on Metric Enrollment
Parame
ters
Algorit
hms
Th
ro
ug
hp
ut
Faul
t
Tol
era
nce
Res
pon
se
Tim
e
Ove
rhe
ad
Res
ourc
e
Utili
zati
on
Scal
abili
ty
Perfo
rman
ce
Round
Robin
YES NO YES YES YES YES YES
ESCE NO NO NO NO YES YES NO
Min YES NO YES YES YES NO YES
Max YES NO YES YES YES NO YES
ALO YES NO NO YES YES NO NO
Honey
Bee
YE
S
NO NO NO YES NO NO
Throttle
d
NO YES YES NO YES YES YES
6. CONCLUSION:
In this paper, we analysed different algorithms in cloud
computing for load balancing. Cloud computing has
generally been implemented by the industry, through
there are many types of existing issues like Server
Consolidated, Energy Management, Load balancing,
Virtual machine Migration, etc. The main issue in all of
these is load balancing, that is necessary to allocate the
excess dynamic local workload equally to all the nodes in
the entire cloud to get higher customer fulfilment and
resource consumption proportion. In this paper, we have
analysed and equated different dynamic and static load
balancing algorithms in cloud computing such as, Max-
Min, Ant Colony Optimization Algorithm, round robin,
Honeybee, Min-Min, Throttled Algorithm etc. considering
the features like overhead, fault tolerance, throughput,
scalability etc.
7. REFERENCES
[1] R. Shimon ski, Windows 2000 And Windows Server
2003, Clustering and Load Balancing Emeryville,
McGraw-Hill Professional Publishing, CA, USA, 2003.
[2] R. Mata-Toledo, and P. Gupta, “Green data centre:
how green can we perform”, Journal of Technology
Research, Academic and Business Research Institute, Vol.
2, No. 1, May 2010.
[3] Ali M Alakeel, “A Guide to Dynamic Load Balancing in
Distributed Computer Systems”, International Journal of
Computer Science and Network Security, Vol. 10 No. 6,
June 2010.

[4] Parin. V. Patel, Hitesh. D. Patel, Pinal. J. Patel, “A
Survey on Load Balancing in Cloud Computing” IJERT,
Vol. 1, Issue 9, November 2012.
[5] Sahu, Yatendra and Pateriya, RK, “Cloud Computing
Overview with Load Balancing Techniques”,
International Journal of Computer Applications,
2013,vol. 65, Sahu2013.
[6] 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 Distributed Computing, Elsevier, Vol. 70, No.
6, May 2010, pages 1-18.
[7] O. Elzeki, M. Reshad, M. Elsoud, “Improved max-min
algorithm in cloud computing, International Journal of
Computer Applications” vol 50 (12) (2012)pages 22–27..

LOAD BALANCING IN CLOUD COMPUTING

More Related Content

Similar to LOAD BALANCING IN CLOUD COMPUTING (20)

More from IRJET Journal (20)

Recently uploaded (20)

LOAD BALANCING IN CLOUD COMPUTING