Load Balancing in Cloud using Modified Genetic Algorithm

Load balancing in Cloud using modified genetic algorithm
Manmohan Sharma 1
, Anil Kumar 2
1
Mody University Of Science and Technology, Laxmangarh, Rajasthan, India
manmohan.manu@gmail.com, dahiyaanil@yahoo.com
Abstract: Cloud computing is a mix of distributed, grid and parallel processing. It is as of late in pattern on account of the
benefits it gives. It gives a pool of resources which are shared among different clients. Alongside its expanding request, it endures
with a few issues. A standout amongst the most vital and testing issue of cloud computing is load balancing. Load balancing
essentially intends to adjust the load similarly among a few hubs so hub is over-burden, under loaded or sitting inactive. Till date
there are numerous calculations proposed to deal with load balancing yet none of them has been demonstrated as productive one.
In this paper a load balancing algorithm is proposed utilizing rule of genetic algorithm. Fitness of assignments is ascertained and
on the premise of fitness load balancing is done. In this algorithm priority is appointed to the wellness computed in like manner
the chromosome with most noteworthy fitness is doled out least priority. Fitness here stands for the aggregate cost needs to
actualize an errand. Increasingly the cost more is the fitness. The entire simulation is performed on cloudsim 3.0 toolbox which is
JAVA based simulator.
Keywords: Cloud computing, Load balancing, Genetic algorithms, Priority, Fitness, Chromosome.
I.INTRODUCTION
Cloud computing is the most recent IT innovation
which is adjusted by various associations on
accounting of its few components it gives. It is
another worldview which gives on-request access to
different resources like storage, compute and
network. These registering administrations are given
by various cloud specialist organizations like
Amazon, Google, and Microsoft [1]. Alongside its
expanding selection, cloud computing suffers from a
considerable measure of genuine difficulties. Load
balancing is one of the genuine trials of cloud
computing which corrupts its execution. It disperses
the workload among nodes so that the no hub is over-
burden, underloaded or sitting inert. To deal with the
issue of load balancing numerous algorithms till date
have been proposed like FCFS, Round Robin,
Honey-Bee foraging and so forth. In any case, none
of the algorithms has tackled this issue totally and
productively. Every procedure set up together by a
customer is viewed as a collection of the assignment
which is managed by nodes by setting up for them to
upgrade execution. The workload is leveled among
different hubs anytime[2]. Genetic Algorithm is an
approach to deal with the load in the framework. It is
an inquiry calculation which utilizes the idea of
hereditary qualities and regular advancement. It
utilizes encounters from the past in future. It utilizes
Darwin theory of survival of the fittest which lets just
the hubs which are fit to stay and deliver assist
Posterity. The paper is sorted out in the
accompanying ways. The Introduction is expressed in
area 1; in segment 2 load balancing is talked about in
a word; genetic algorithm is expressed in segment 3;
Load balancing utilizing GA has been proposed in
segment 4; in segment 5 simulation environments is
discussed; in segment 6 algorithm is proposed. At
last, concluding remarks show up in area 5.
2. LOAD BALANCING
Exactly when something is talked about on load
balancing, then it has been said that load balancing is
known as a system which spreads the workload of a
particular hub to all other neighboring hubs to make
the hub work speedier and the goal is to limit the
general execution time. In this technique, it will be
thought to have a high customer satisfaction and
resource usage part and it will be made an indicate
have no single hub is wary, that the general
presentation of the affiliation will be certainly
extended. It too supports in applying flop over,
permitting versatility, keep up a key separation from
a couple of drawbacks like the bottlenecks and over
provisioning, sinking response time et cetera. As this
has starting at now been discussed Load balancing
has its guideline focus to pass on the load between
hubs or we can state between different resources of
an affiliation. Cloud specialist co-op is reliant on a
few instruments of programmed load balancing, that
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 9, September 2017
1 https://sites.google.com/site/ijcsis/
ISSN 1947-5500

is with the change of requests, and the clients will
build the no. of CPUs for their resources. These all
fields are constantly reliant on the client business
prerequisites. So there are two basic needs of load
balancing are, first is to support accessibility of
Cloud resources and second is to reinforce execution.
Some critical objectives of load balancing are:
a) Cost effectiveness: The load balancing algorithm
ought to be cost productive. The general cost to
execute the calculation ought to be in evaluated
spending plan.
b) Elasticity and Scalability: The connected load
balancing algorithms ought to be versatile to
alterations. Henceforth it will bolster issues like
adaptability and flexibility.
c) Priority: It is the most essential idea utilized as a
part of load balancing algorithms. They ought to do
need of the resources accessible with a specific end
goal to enhance general productivity. Prioritization
will prompt to less execution time.
3. GENETIC ALGORITHM
A genetic algorithm is a guideline of delicate
processing which depends on the idea of common
hereditary qualities and advancement. It depends on
Darwin's hypothesis of survival of the fittest which
lets just the fits hubs to stay and live and to create
better posterity's which in regard builds the general
execution. The genetic algorithm comprises of strings
which are artificial creatures and utilizations the data
of past strings in each new era. A genetic algorithm is
arbitrary yet despite everything it utilizes chronicled
data for better outcomes. A genetic algorithm is
extremely basic; it duplicates the string and halfway
or more than mostly swaps them in light of the fitness
[8][9]. This entire procedure is conveyed by 3
operations: selection, crossover and mutation.
Selection depends on survival of fittest rule and
chooses just the hubs which are fit and disposes of
the rest. There are distinctive approaches to the
selection procedure which incorporates roulette
wheel, tournament algorithm and so forth. Next 2
operations are in charge of investigating new
components [10][11]. Crossover trades partitions
between strings. There are distinctive approaches to
play out this operation like single point crossover,
multi-point crossover and so forth. The outcome
delivered from this operation is named as children.
To change the qualities of a chromosome from a
characterized one mutation operation is conveyed.
Change is not conjured dependably; it relies on upon
mutation likelihood. What's more, contingent on that
the bits of chromosomes are flipped from 0 to 1 or 1
to 0. This procedure is conveyed till the fittest
chromosomes are not accomplished[12][13].
Total Number
of jobs
Too many
>M2
Moderate
>M1 and <=M2
Very less
<=M1
Processing rateProcessing rateLightly loaded
Rate=LowRate=High Rate=High Rate=Low
Heavily loadedCan handle
jobs
Lightly loaded Scheduling
Figure 1 Decision tree on VM’s
ISSN 1947-5500

4. LOAD BALANCING USING GA
In spite of the fact that cloud computing is rapid in
nature, however at a particular occurrence load
balancing is defined as allotting N assortment of jobs
presented by cloud clients to M assortment of
handling units inside the cloud[13]. Each of these
units will have a vector showing the amount CPU has
been used. This vector comprises of MIPS (a large
number of instructions every second), C, cost of
execution and D, latency cost[6][7]. The delay value
gauges penalty which demonstrates how much cloud
specialist organization must pay to the customer if
the job is not finished inside the mentioned days.
JP= f (MIPS, C, D) (1)
Unit of job (JU) consists of a job submitted by user.
UOJ= f (st, N, AT, MT) (2)
Where, s speak to type of service required by the
occupation SAAS, PAAS or IAAS. N speaks to an
aggregate number of instructions in a job which
should be executed, AT alludes to the arrival time of
a job, MT alludes to the total time required to finish
the job.
The cloud service providers need to allot these K jobs
among M number of processors with the end goal
that the estimation of cost capacity (CF) is limited.
CF= w1*C(N/MIPS) +w2*D (3)
Where w1 and w2 are predefined weights that are 0.8
and 0.2 respectively such that value of their
simulation is always 1.
A load of each virtual machine can be calculated by
using the below equation:
Li=np
+ nq
+ nr
(4)
5. SIMULATION ENVIRONMENT
Simulation remains for making a domain which looks
and carries on like unique one. The proposed
procedure or algorithm can be examined in a
simulation environment. By utilizing the
effectiveness, execution and so forth can be
examined. It is more profitable to clients since they
can watch their algorithm before executing it on real
condition. Additionally, it decreases general cost as
changes can be made before acknowledging it
really[3]. The entire situation will go to execute in
cloudsim 3.0.[4][5] shown in figure 2.
Distinctive functionalities of cloudsim 3.0 are:
1.Support simulation of huge scale cloud computing
data centers.
2.Support simulation and demonstrating of virtual
server host.
3.Support simulation and demonstrating of
computational resources.
4.Support simulation and demonstrating of
datacenters.
5. Support for user defined scheduling policies.
Cloudsim
Datacenter
Broker
Datacenter
Characteristics
Datacenter
VMAllocation
Policy
Network
Topology
SAN Storage
Cloudlet
VMAllocation
PolicySimple Federated
Datacenter Cloud
Cordinator
Sensor
VM Cloudlet Scheduler
RamProvisonerHost
BwProvisoner
CloudletScheduler
TimeShared
CloudletScheduler
SpaceShared
RamProvisoner
SimpleVMScheduler
BwProvisoner
Simple
VMScheduler
SpaceShared
VMScheduler
TimeShared
Figure 2 Cloudsim 3.0
ISSN 1947-5500

6. PROPOSED ALGORITHM
Step 1: Obtain data about VM's from datacenters and
store them.
Step 2: Assign weightage to VM's on the premise of
parameters like storage space, MIPS, processor and
so on like if a VM can deal with a twice as load as
another it will be weighted "2" or in the event that it
can deal with load 4 times as another it will be
weighted '4'.
Step 3: Calculate a load of each virtual machine by
utilizing condition 4 and settle on a decision tree on
the premise of that as showed in figure 2.
Step 4: Check if the load is balanced or if there is any
overloaded node. If yes, then balance the load.
Step 5: Load tasks of the cloud users incorporating
parameters showed in condition 2.
Step 6: Convert these jobs into binary strings called
chromosome.
Step 7: Calculate fitness of every chromosome by
utilizing condition 3.
Step 8: Eliminate chromosome with high fitness
condition (Selection).
Step 9: Perform single point crossover form new
offspring matched with the index of VM (crossover).
Step 10: Perform mutation with probability 0.5.
Step 11: Add the new chromosomes to the present
population and assess fitness. If fit, go to step 12 else
repeat step 8 to 11 again and again till desired fitness
is achieved.
Step 12: Arrange the chromosomes in increasing
order of fitness.
Figure 3
Start
Collect information
of VM's
Assign weightage to
VM's
Make decision tree
Load tasks of
clients
Convert jobs into
binary strings
Calculate fitness of
each VM
Selection
Mutation
Add new off springs to
current population
Evaluate fitness
Fit
Perform sorting
Crossover Generate new Decision
Tree
Is load balanced
Failure of VMAre VM
overloaded
Transfer load
No Yes Yes No
End
Schedule tasks
Not fit
ISSN 1947-5500

Step 13: Schedule them to the VM's all together of
their fitness as per Step 12 and update the status of
the VM's
Step 14: Create another decision tree as per the
present status of the VM.
Step 15: The load will be adjusted in two cases:
Step 15 (a): Calculate load of each VM in the
decision tree utilizing condition 1. If any load is
heavily loaded or couldn’t handle the load, perform
load balancing of that VM using steps 7 to 15.
Step 15 (b): Check if there is a failure in any of the
VM. If yes, transfer the load of that VM to another
one.
The entire proposed algorithm is exhibited in figure
3.
The above expressed situation is of genetic algorithm
utilized as a part of load balancing. Presently we will
examine about how the errand will be planned for the
virtual machines. Now we will discuss about how the
task will be scheduled in the virtual machines.
Step 1: The task will be alloted to the VM scheduler
which thus will check the VM to which the work can
be allocated by searching for it in the decision tree.
Step 2: Once the scheduler finds a reasonable VM to
dole out the job, it will predict the type of service
required stated in equation 2 and will calculate the
capacity of the VM required by using equation 5
which will contain of 3 parameters.
VMc = (Number of available VMs) * (utilization) *
(effieciency) (5)
Where VMc stands for the capacity of the VM.
Step3: After that VM will compute the make span
time of that specific task by utilizing condition 6 and
will compare it with estimated evaluated time. If it is
less than the estimated time the task will be
forwarded to the VM to execute else it will be
discarded.
MST= (Task length/VMc) + WT (6)
Where MST and WT remains for make span time and
waiting time individually. Task length will be
calculated by calculating number of bits in
chromosome. The entire situation is expressed in
figure 4.
6. CONCLUSION AND FUTURE WORK
In this paper choice tree based hereditary calculation
has been proposed to perform stack adjusting on hubs
and to deal with hub disappointment. By utilizing
hereditary calculation correspondence cost, reaction
time and so forth are limited and the general
execution is expanded[14]. The principle objective of
utilizing this calculation is to adjust the heap in the
framework viably and effectively. Need is doled out
to the occupations of clients by figuring the wellness
work which exhibits the aggregate cost of that of a
Figure 4 Task scheduling in VM
specific assignment. Employments with high
wellness work that is high cost are wiped out in the
determination strategy. The entire system is
additionally expressed utilizing flowchart. In future
the work can be reached out by utilizing make
traverse, cost and MIPS joined as wellness capacity.
Additionally variety of hybrid and change procedures
should be possible for expanding execution and
productivity.
Task
Vm scheduler
Find VM and
Calclaute VMc
Predict Service
Required
Estimate time of task processing
If (actual time <= estimated time)
NoYes
Forward the task to
VM
Process task Reject task
ISSN 1947-5500

REFERNCES
[1] Amandeep, “Analysis of Load Balancing
Techniques in Cloud Computing”, International
Journal of Computers & Technology, vol. 4,
no.2, pp.2277-3061, March-April, 2013.
[2] Xiaocheng Lui,Cheng Wang,Juliang Chen and
Albert Y.Zomaya,”Priority-Based Consolidation
of Parallel Workloads in the Cloud”, IEEE
Transactions On Parallel And Distributed
Systems,vol.24, no.9, pp.1874-1882,September,
2013.
[3] Luiz F.Bittencourt,Edmundo R.M.Madeira, and
Nelson L.S da Fonseaca,”Scheduling in Hybrid
Clouds”, IEEE Communications Magazine,
vol.4, pp.42-47, September,2012.
[4] B. Wickremasinghe, R.N. Calheiros and R.
Buyya, “Cloudanalyst: A cloudsim-based visual
modeller for analysing cloud computing environ-
ments and applications”, in Proc. of Proceedings
of the 24th International Conference on
Advanced Information Networking and
Applications (AINA 2010), Perth, Australia,
pp.446-452, 2010.
[5] R.N. Calheiros, R. Ranjan, A. Beloglazov, C.
Rose, R. Buyya, “Cloudsim:A toolkit for
modeling and simulation of cloud computing
environments and evaluation of resource
provisioning algorithms”,in Software:Practice
andExperience(SPE),Vol:41,No:1,ISSN:003806
44,Wiley Press,USA,pp:23-50,2011.
[6] Nikravan, M. and Kashani, M.H. “A Genetic
Algorithm For Process Scheduling In Distributed
Operating Systems Considering Load balancing”
in Proceedings of the 21th European Conference
on Modeling and Simulation, 645-650, 2007.
[7] A. Y. Zomaya, & Y. H. The. “Observations on
using genetic algorithms for dynamic load
balancing” IEEE Transactions on Parallel and
Distributed Systems, pp. 899-911, 2013.
[8] T. Lim and H. Haron. “Performance comparison
of genetic algorithm, differential evolution and
particle swarm optimization towards benchmark
functions” Proc. de IEEE conference on open
systems (icos), Kunching, pp. 211-215, 2013.
[9] D. Yuan, Y. Yang, X. Liu, W. Li, L. Cui, M. Xu,
J. Chen. “A Highly Practical Approach towards
Achieving Minimum Datasets Storage Cost in
the Cloud. IEEE Transactions on Parallel and
Distributed Systems, pp. 1234-1244, 2013.
[10] S. Singh and M. Kalra, "Scheduling of
Independent Tasks in Cloud Computing Using
Modified Genetic Algorithm," in Computational
Intelligence and Communication Networks
(CICN), 2014 International Conference on, pp.
565-569, 2014.
[11] J. W. Ge and Y. S. Yuan, "Research of cloud
computing task scheduling algorithm based on
improved genetic algorithm," in Applied
Mechanics and Materials, pp. 2426-2429, 2013.
[12] R. Kaur and S. Kinger, "Enhanced Genetic
Algorithm based Task Scheduling in Cloud
Computing," International Journal of Computer
Applications, vol. 101, 2014.
[13] T. Goyal and A. Agrawal, "Host Scheduling
Algorithm Using Genetic Algorithm in Cloud
Computing Environment," International Journal
of Research in Engineering & Technology
(IJRET) Vol, vol. 1, 2013.
[14] Z. Zheng, R. Wang, H. Zhong, and X. Zhang,
"An approach for cloud resource scheduling
based on Parallel Genetic Algorithm," in
Computer Research and Development
(ICCRD), 3rd International Conference , pp.
444-447, 2011.
ISSN 1947-5500

Load Balancing in Cloud using Modified Genetic Algorithm

Recommended

More Related Content

What's hot (19)

Similar to Load Balancing in Cloud using Modified Genetic Algorithm (20)

Recently uploaded (20)

Load Balancing in Cloud using Modified Genetic Algorithm