Sie sind auf Seite 1von 5

2018 8th International Conference on Communication Systems and Network Technologies

Virtual machine selection strategy based on Grey Wolf Optimizer in Cloud


Environment: A study
Priyanshi Singh Murtaza Abbas Rizvi
Computer Engineering and Applications, Computer Engineering and Applications,
NITTTR, Bhopal, India NITTTR, Bhopal, India
Priya92singh@hotmail.com marizvi@nitttrbpl.ac.in

Abstract - In cloud computing environment there are many


strategies used for virtual machine (VM) placement. B.BENITA JACINTH AND ET AL.[1] - In this paper,
Virtualization technology can create multiple virtual authors projected a multi-objective hybrid ACO-PSO
machines. The mapping method of virtual machines to improvement algorithmic program for minimizing resource
physical machines is thought as virtual machine placement.
wastage, minimizing the wastage of power and resource by
Objectives for VM selection can be multi-objective and mono
objective. MP’s main aims are to reduce the make span time applying intelligent load equalization policy in physical
and speed. Cloud computing plays a vital role in content servers. Simulation results show that the algorithmic
delivery services for processing data and scheduling the task program reduces resource wastage and power consumption
like smart cities sensors data, healthcare data processing and additionally provides load equalization in servers in
services (HDPS). Cloud computing for HDPS can restore comparison to the prevailing multi-objective ant colony
patients’ records, diseases diagnosing and different medical system algorithmic program.
domains in less time and fewer of price. Here we proposed a
methodology which is GWO algorithm based on nature MD. MOFIJUL ISLAM, MD. ABDUR RAZZAQUE
inspired concepts. The algorithm will work on optimal ET AL.[7] In this paper, author tend to propose an ant
selection of virtual machine. Colony optimization (ACO) primarily based joint VM
migration model for a heterogeneous, Mobile Cloud
Keywords: Cloud computing, HDPS, Virtual machine, Computing primarily based enhanced based enhanced
GWO. healthcare data processing system for smart city
environment.. During this model, the users quality and
I. INTRODUCTION provisioned VM resources within the cloud address the VM
In cloud computing services(SAAS,IAAS,PAAS) the migration disadvantage. Author additionally present a
selection of optimal and most suitable virtual machines thorough performance analysis to analyze the effectiveness
(VMs)for high computation delivery services like of the projected model compared to progressive
healthcare services (HCS) where data delivery and task approaches.
computation on time is extremely important .The task T.THIRUVENKADAM ET AL [4] In this paper author
includes record of patients, doctors, etc. In this situation, proposed optimized algorithms for virtual machines
cloud computing (CC) will play a significant role by placement based on multi objective characterstics.Paper is
dealing with massive health care data related tasks like composed of two main task scheduling and load balancing.
sharing, processing and analyzing, Cloud computing work
on dynamics provisioning concept. In dynamic provisioning XIONG FU AND CHEN ZHOU ET AL [15]- In this
services, network, storage, consumption, management and paper the author described concept of affinity based VMP
may infrastructure services comes to fulfil the required of in cloud environment. Author proposed a model, which
user and getting benefits to cloud service and providers. evaluates the similitudes between virtual machine and
Utilizing dynamic useful resource allocating for balancing host.VM will be placed on those host who have highest
loads and mapping the requirement is very important. In affinity advantages.
order to achieve maximum efficiency and scalability in a NILESH PACHORKAR1, RAJESH INGLE ET AL
fact manner provider should balance loads with efficient [13]- In this paper author proposed a multidimensional
and optimal scheduling algorithms. affinity model which uses multiple dimensions, such as
CPU usage, memory, bandwidth etc. The proposed
II. RELATED WORK algorithm showed optimized performance as compared to
Through related work, on ideal arrangement determination others.
and distribution approaches the same number of kinds of TEVFIK YAPICIOGLU, SEMA OKTUG [10]- In this
research were done on applying and utilizing astute paper author considered network traffic as an input for
calculations and algorithms such as PSO, FFO,GA, ACO placement problem. VMs are clustered based upon traffic
and Swarm intelligent. But having some drawbacks like networked. Thus proposed algorithm shows quick and
local optimum entrap global updating problems and less efficient results to improve the placement.
convergence make most of the algorithms not so feasible.

978-1-5386-5956-4/18/$31.00 ©2018 IEEE 108


DOI: 10.1109/CSNT.2018.23
III. PROPOSED INTELLIGENT ARCHITECTURE Here once our improvement approach ids outlined then
OF CLOUD FOR HCS article for analysis may also be classified because the
objective perform. Then the target perform is optimized
With the increasing number of the smart city; smart
individually or is optimized at the same time.
healthcare services are rising to enhance urban citizens’
quality of life [6][7][8], however to access and process
smart data processing services are an important issue for Classification
the people with sensors seamlessly anywhere and anytime. criterion:
Cloud computing is also plays an important role in Mobile
CC services, users will submit a task in cloudlet for faster
execution. Existing literature discusses many
methodologies to observe how fast and optimized task get
scheduling and submission in a cloudlet.
A network administrator is responsible for applying the
intelligent technique (GWO algorithm) .That it uses to get
the optimum chosen of VMs within the cloud .So to
enhance the task (the main aim of task scheduling is to
improve the performance and quality of service) process
that results in total time minimization for HDPS requests
and maximize resources utilization.

FIGURE 2 Main Classification Criterion for VMP/S Taxonomy

Hence an appropriate improvement method has ought to be


adopted to determined that how the output of operational
optimized unit could be done. now day meta heuristic
improvement methods became the a lot of popular over the
past years thanks to having some points like-simple,
derivation free and most significantly exclusion of local
optimum.
Meta heuristic technique will have divided supported –
x Single solution based
x Population based
Swarm intelligent behavior belongs to population based
Meta heuristic. Some swarm intelligence algorithms are-
Marriage in Honey Bees improvement algorithmic program
(MBO), Artificial Fish-Swarm algorithmic program
FIGURE 1.Architecture of Cloud Computing For HDPS (AFSA), Bird mating Optimizer (BMO), Firefly algorithm
(FA) ,Particle swarm optimization, ant colony optimization,
IV. PROPOSED GWO BASED ALGORITHMS FOR Glow warm optimization, firefly optimization ,Fruit fly
CLOUD COMPUTING ENVIRONMENT optimization algorithm (FOA) etc.

A. Background B. GWO (Grey Wolf Optimization)


Virtual Machine allocation or choice is a problem of np GWO is a swarm intelligence technique developed by
hard. This operational drawback is formulated or you will Mirjaili et al, Which imitate the hierarchy in leadership of
say divided in dynamic and static environment. The wolves and they are known for their group hunting. Grey
fundamental taxonomy of Virtual machine placement is Wolf Optimizer is a bioinspired algorithm. It is basically
sections in three ways in which [16] simulate the behavior of grey wolf. As grey wolf shown
1. Optimization approach social hierarchy of dominant. Grey wolves have many
2. Objective function approach strong abilities like preying, encircling and also memorizing
3. Solution techniques the position of prey. GWO belongs to family Canidae
Basically optimization approach will have suggested based family and largely choose to live in a pack. gray wolfs are
on several methodology like: considered as apex predators. within the alternative words
1. Mono objective we will say that they're at prime of food chain. In general
the cluster size is 5-12 on the average while they are on
2. Multi objective
food or run.
3. Pure multi objective

109
mathematically model approach the prey price a Ԧ need to
C. Mathematical Model And Algorithm be decrease, so purposeful vary of A is in addition reduced
Here mathematical modeled is framed via social hierarchy, by a Ԧ .Value of a And A parameters can be help to find
tracking, encircling and attacking prays that lined the GWO global minimum with fast convergence speed. Parameter a
algorithm’s define. is basically used here to control the tradeoff between
Social Hierarchy- exploration (which improves diversity of solution) and
As we all know that GWO {is primarily based} on exploitation (which improves quality of solution).It is
population based nature inspired algorithmic program. This linearly updated from 2 to 0 in each iteration.
algorithmic program contains a few parameters solely and
simple to implement that build it higher than alternative 2
a= 2− t . −1
one. So we are going to think about the fittest resolution as MaxIter
alpha (α) and alternative resolution are β,δ. D is enriching (IV)
pray.
Encircling Pray- E. PROPOSED METHODOLOGY AND
The mathematical model of the peripheral behavior [23] – ALGORITHM
The Pseudo code for GWO algorithmic program is as
D= ¿C
|Ԧ X p (t )− . Ԧ
Ԧ. Ԧ X (t )‫¿ ש‬ (1) follows-

X (t+1)= Ԧ
Ԧ X p (t )− Ԧ
A .Ԧ
D (2)
Where-
t=current iteration
A,C are coefficient vector
Xp=position vector of pray
Ԧ
X=position of grey wolf

Calculation for A, C as follow

A= 2Ԧ
Ԧ a .x 1− Ԧ
a (3)

C= 2. x2
Ԧ (4)

Where value of a is linearly decreased from 2 to 0 over the


iteration course. Also x1, x2 are random vectors
[0,1].Vector C is a random value whose range is [0,2].Grey
wolves have ability to acknowledge the location of prey and FIGURE.3 GWO Flow Chart
encircle them. We use following formulas-[GWO]
Our paper design using GWO in this section, the pattern
ԦDα= ¿Ԧ
C 1.ԦXα− Ԧ
X‫¿ ש‬ (5) which is used to design the optimal .This paper considering
Ԧ Ԧ
Dβ= ¿C 2. Xβ− X ‫¿ ש‬
Ԧ Ԧ (6)
the fitness function .Optimal determination of the proposed
Algorithm’s parameters is transformed to 0.
Ԧ C 3. Ԧ
Dδ= ¿Ԧ Xδ− Ԧ
X‫¿ ש‬ (7) Here some points for to show how GWO is in theory able to
X(t+1)=X1+X2+X3/3 (8) solve improvement issues:
Value of X can be determined by using following 1. Here the methodology for GWO is save the best
equations- solution when applied the course of iteration.

X 1= Ԧ
Ԧ Xα− Ԧ
A 1.(Ԧ
Dα) 2. As value of A can be separated as |A|≥ 1 shows
(I)
exploration and |A|<1 shows exploitation.
X 2= Ԧ
Ԧ A 1.(Ԧ
Xβ− Ԧ Dβ) (II)
3. In the last, parameter A and C should be adjusted
according to need and iteration.
X 3= Ԧ
Ԧ A3.(Ԧ
Xδ− Ԧ Dδ) (III)

Here Xα, Xβ, Xδ are the best three solution of wolfs. Also
A1,A2,A3 can be calculated from equation (3), so as to

110
TABLE 1. PARAMETER OF VMS OPTIMAL SELECTION In order to illustrate our work, assume that we need to show
SN DATA DESCRIPTION relationships between makespan (time)of request from any
application (HDPS)and number of processors. The results
shows that, whenever the increased the number of
The percentage of CPU capacity
used During a specific period of
processors that reflects the decrease within the makespan.
1 CPU Utilization (U) time.

Time difference between


Turnaround time completion time and arrival
2 (TT) time.

Time difference between


3 Waiting time (WT) turnaround time and burst time.

A. Experimental Results
This section basically composed of the results of
implementation. The PSO and GWO algorithms are
implemented on Cloudsim .The Results Shows as follow-

B. Implementation of GWO Algorithm on CloudSim


In this section we propose GWO algorithm on cloudsim to
get optimal chosen of VM’s for HDPS to minimize the
implementation time for generated requests. Also here we
get results on maximization of resource utilization.
Here we target results based on default CloudSim which
work sequentially like first task will take first VM and FIGURE 6. The proposed GWO algorithm leads to minimized makespan
second will take second VM and so on. Results shows total
So the result shows that GWO algorithm performs better
time for build successful cloudlet is 4 sec as shows in fig, 4
then PSO in terms of makespan. So we can understand that
makespan time is decreasing when applied GWO algorithm.

5. Conclusion
As this paper work also introduced enhanced intelligent
architecture based on smart cities work ,content delivery
services over cloud and healthcare services which use
GWO rule to urge best chosen of VMs in a cloud
environment supported SAAS architecture. This paper
showed GWO algorithm which is a population based meta-
FIGURE 4. Results based on default CloudSim heuristics algorithm better than GA and PSO to
implementation time, more efficient and have showed
Now we target results based on GWO algorithm in minimum makespan. Therefore, this paper planned and
CloudSim which work on the basis of how the Grey wolf analyzed GWO algorithm that is a bio inspired algorithm to
algorithm will work so on. Results shows total time for get the most effective and optimized chosen of VMs to
build successful cloudlet is 1 sec as shows in fig, 5 support Content process Services (Healthcare processing
Services) in minimizing implementation time of medical
requests (tasks), best manage task planning, maximize
resources utilization as applied this algorithm on CloudSim
package simulator. The future work to get an optimal
choice of VMs on a cloud environment with energy
efficiency.

REFERENCES
[1]. B. Benita Jacinth Suseela1, V. Jeyakrishnan, “A Multi-Objective
Hybrid ACO-PSO Optimization Algorithm For Virtual Machine
FIGURE 5. Results based on GWO algorithm in CloudSim Placement In Cloud Computing,” IJRET: International Journal of

111
Research in Engineering and Technology, Vol.3, No.4, pp.474-476,
Apr 2014.
[2]. A. C. Adamuthe, R. M. Pandharpatte, G.T. Thampi, “Multi
objective Virtual Machine Placement in Cloud Environment”,
International Conference on Cloud & Ubiquitous Computing &
Emerging Technologies, pp.1-4, 2013.
[1] Jinchuan Han, Yu Liu, Gang Wu, “Cloud Computing Virtual
Machine Allocation Strategy Based on Social Force Swarm
Intelligence Optimization Algorithm”, Revista de la Facultad de
Ingeniería U.C.V., Vol. 32, No.10, pp. 806-813, 2017.
[2] T. Thiruvenkada, P. Kamalakkannan “Optimized Algorithms for
Virtual Machine Placement based on Multi-Dimensional Resource
Characteristics in Cloud Computing Systems,” International Journal
of Computational Intelligence and Informatics, Vol. 6: No. 1,
pp.53-65, June2016.
[3] Foudil Abdessamia1,Yu Tai2, Wei Zhe Zhang1, Muhammad
Shafiq, “An Improved Particle Swarm Optimization For Energy-
Efficiency Virtual Machine Placement”,5th International
Conference on Cloud Computing Research and Innovation. Pp.1-5,
Apr 2017.
[4] Jemal Hanen,Mounir Ben Ayed, “An enhanced healthcare system in
mobile cloud computing Environment”, Vietnam Journal of
Computer Science, Vol.3, No.4, pp.267-277, July 2016.
[5] Md. Mofijul islam,Mobile “Cloud-Based Big Healthcare Data
Processing in Smart Cities,” .IEEE Access, Vol.5, pp. 11887 –
11899, May 2017.
[6] Ahmed Abdelazziz, Md. Elhoseny, Ahmed S Salama, A.M. Riad,
A.E. Hassanien, “Intelligent Algorithms for Optimal Selection of
Virtual Machine in Cloud Environment, Towards Enhance
Healthcare Services”, International Conference on Advanced
Intelligent Systems and Informatics, pp.289-298, Aug 2017.
[7] Renu Saini, Nishant Anand, “A Multi-Objective Ant Colony
System Algorithm for Virtual Machine Placement”. Journal of
Engineering Research and Application, Vol. 7, Issue 1, ( Part -4)
pp.95-97, January 2017.
[8] Tevfik Yapicioglu, Sema Oktug, “A Traffic-Aware Virtual
Machine Placement Method for Cloud Data Centers,” IEEE/ACM
6th International Conference Utility and Cloud Computing, pp.1-4,
2013.
[9] Pooja Daharwal, Varsha Sharma, “Energy Efficient Cloud
Computing VM Placement Based On Genetic Algorithm,”
International Journal of Computer Trends and Technology (IJCTT),
Volume 44 Issue 1, pp.15-23, February 2017
[10] Xin Ye, Yanli Yin, Lan Lan, “Energy-Efficient Many-objective
Virtual Machine Placement Optimization in a Cloud Computing
Environment”, IEEE Access, Vol.5, pp.16006-16020, 2017.
[11] Nilesh Pachorkar, Rajesh Ingle, “Multi-dimensional Affinity Aware
VM Placement Algorithm in Cloud Computing,” International
Journal of Advanced Computer Research, Volume. 3, No. 4, Issue-
13, December 2013.
[12] Dilip Kumar, Tarni Mandal, “Multi-Objective Virtual Machine
Placement using Improved Teaching Learning Based Optimization
in Cloud Data Centers,” International Journal of Applied
Engineering Research, Volume 12, Number 21 (2017).
[13] Xiong Fu and Chen Zhou, “Predicted Affinity Based Virtual
Machine Placement in Cloud Computing Environments,” IEEE
Transactions on Cloud Computing”, Early Access, 2017.

112

Das könnte Ihnen auch gefallen