7
A Technical Review on Cloudsim based VM Scheduling Techniques in Cloud Computing Environment Nagadevi.S SRM University Chennai, India Dr.S.V.Kasmir Raja Dean, Research SRM University Chennai, India AbstractVirtualization is a key concept of cloud computing. Data centers virtualizes their physical resources to improve resource utilization, revenue maximization ,power consumption and so on. Virtual machines are created as separate entity to run users applications. To make effective use of these virtual machines an optimized virtual machine placement decision has to be made. Placing VMs on suitable PMs is known a virtual machine placement. This paper analyses various algorithms and techniques used for virtual machine placement. Index TermsVM provisioning , allocation, auction-based I. Introduction Nowadays, organizations move their business to their own Datacenters. Datacenters provide the environment for the user’s applications to run. Efficient utilization of Datacenter resources is the key issue for every organization. With the help of virtualization concept, organizations virtualize their physical recourses to service many request to run an application. The Virtual Machines (VM) are created and assigned to them by the Cloud Service Provider. The key challenges faced by the organizations, in maintaining the datacenters are: how to reduce the power consumed by unused (idle) resources, how to dynamically schedule the VMs among the users application to improve multidimensional resource usage. II. Background When a user submits the job to the Datacenter , the job scheduler will configure appropriate resources i.e. VMs. i.e. scheduling jobs onto virtual machines is done in PAAS. Job scheduling includes accepting or rejecting the request based on the available VMs, and also optimizing the mapping of jobs to VMs. Once the requested VMs are configured and mapped to jobs, these VMs are actually created in a physical server called Physical Machine .So, scheduling VMs onto physical machines is known International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com Page 6 of 12

A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

A Technical Review on Cloudsim based VM Scheduling Techniques in

Cloud Computing Environment

Nagadevi.S

SRM University

Chennai, India

Dr.S.V.Kasmir Raja

Dean, Research

SRM University

Chennai, India

Abstract— Virtualization is a key concept of

cloud computing. Data centers virtualizes their

physical resources to improve resource

utilization, revenue maximization ,power

consumption and so on. Virtual machines are

created as separate entity to run users

applications. To make effective use of these

virtual machines an optimized virtual machine

placement decision has to be made. Placing VMs

on suitable PMs is known a virtual machine

placement. This paper analyses various

algorithms and techniques used for virtual

machine placement.

Index Terms—VM provisioning , allocation,

auction-based

I. Introduction

Nowadays, organizations move their business to

their own Datacenters. Datacenters provide the

environment for the user’s applications to run.

Efficient utilization of Datacenter resources is the

key issue for every organization. With the help of

virtualization concept, organizations virtualize their

physical recourses to service many request to run an

application. The Virtual Machines (VM) are created

and assigned to them by the Cloud Service

Provider. The key challenges faced by the

organizations, in maintaining the datacenters are:

how to reduce the power consumed by unused (idle)

resources, how to dynamically schedule the VMs

among the users application to improve

multidimensional resource usage.

II. Background

When a user submits the job to the Datacenter , the

job scheduler will configure appropriate resources

i.e. VMs. i.e. scheduling jobs onto virtual machines

is done in PAAS. Job scheduling includes accepting

or rejecting the request based on the available VMs,

and also optimizing the mapping of jobs to VMs.

Once the requested VMs are configured and

mapped to jobs, these VMs are actually created in a

physical server called Physical Machine .So,

scheduling VMs onto physical machines is known

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com

Page 6 of 12

Page 2: A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

as VM scheduling. VM scheduling is done in IAAS.

Allocating virtual machines to physical machines is

one of the key tasks of Infrastructure as a service

(IAAS) cloud. VM scheduling Algorithms are used

to schedule the VM requests to the physical

machines of a data center (DC)

III. VM Placement

VM placement technique is used whenever a user

request for a resource (VM) in a datacenter.

Allocating the requested VM on a PM is a VM

placement technique. VMs should be placed on

PMs such that any VM placement should not

overload any PM and VM placement must reduce

the number of active PMs. The main objective of

the VM placement is to reduce energy consumed by

the datacenter. Energy consumption is reduced by

minimizing the number of active physical machines.

By reducing the number of VM migrations, the

energy, cost and effort of a datacenter could be

greatly reduced.VM placement technique is

required at two instances of a datacenter. One is

initially when the datacenter is unloaded .Another

one is when there is a need to reallocate the VMs in

a dynamic cloud environment.

The VMP problem is selecting a suitable PM to host

a new VM so that the performance of the system is

higher i.e. packing VMs onto a fixed capacity PMs

e.g. Multi Dimensional Bin packing problem.

VM placement decision is required at two places.

One is Initial- Static VM placement which places

VMs at once in an unloaded DC. Second is

Dynamic VM placement which place Vms at run

time to cope up with the dynamic environment.

VM placement decisions are made based on

Reservation, On-demand(Initial VM

placement,Migration of VM),Spot Market.

IV. VMP Problem Definition:

Given a set of n virtual machines and a set of m

physical machines mapping n VMs to m PMs is

called VM placement. Dynamic VM placement

assumes that m PMs are already allocated with n1

number of virtual machines. Now at a point in time

dynamic VM placement involves mapping n+n1

number of VMs to m PMs.

The main objectives of the VMPP are:

1. To reduce energy consumption by reducing

number of running physical machines

2. To do dynamic resource allocation

3. To improve resource utilization

4. To minimize a cost of a data center

5. To improve SLA

6. To reduce number of VM migrations

V. Related Work:

In [1] ,VM Placement decisions were made

based on the following: Reservation, On-

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com

Page 7 of 12

Page 3: A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

Demand and Spot Market. On-Demand VM

placement decisions are based on initial VM

Placement and migration of VMs. Both of the

above steps can be considered for power

conservation, SLA, revenue maximization and

reliability. In this paper only initial VM

placement is studied. Initial VM placement done

at two levels: 1. Cluster / Cloud in which VMs

communicate across cloud. 2. Node/PM in

which VMs communicate only within a single

PM. In VM placement algorithms, physical

machines are partitioned into two sets : those

that meet some criterion(candidates) and those

that do not. Within the candidate PMs , PMs

were ordered based on some heuristics like

online bin packing and then VMs are placed on

PMs until all the PMs are exhausted.

(a) User submits the request to cloud

controller. Request carries number of VMs

of each type (small, medium, Large)

(b) Cloud controller responds with

a. Full grant: allocated the

maximum request

b. Partial grant: allocated instances

less than maximum request

c. NERA :Not Enough resources

Available

(c) On which cluster should requested VMs are

placed based on

a. LLF Least full First

b. PAL Percent Allocated

c. RAN Random

(d) On which node should requested VMs be

placed based on Six Heuristics:

a. FF First Fit

b. LF Least Full First

c. MF Most Full First

d. NF Next fit

e. RA Random

f. TP Tag and Pack

(e) VMs are placed on a selected node on a

selected cluster. If not, the next node in the

list is selected . Node controller reallocates

the VM to the next node in the set. Else

NERA is returned to cloud controller.

In this paper [2]reallocation of VMs performed to

minimize the number of physical nodes. Idle nodes

are switched-off to decrease the power

consumption. VM provisioning and VM placement

were done using Bin packing using MBFD. VMs

are selected for migration based on MBFD.The

objective of this paper is to consolidate VMs

leveraging live migration and switch off idle

nodes to minimize power consumption while

providing QoS .BFD algorithm is used for VM

allocation. The following heuristics are used for

selection of VMs for migration , Single Threshold,

Minimum number of Migrations, Highest Potential

Growth and Random Choice.In this paper they have

considered only single core nodes.

The objective of this paper [3] is to map multi-core

VMs to multi-core PMs using constraint

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com

Page 8 of 12

Page 4: A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

programming. Scheduling of multiple cores in VM

placement leads to optimized results for overall

performance and cost.VM placement is combined

with core scheduling to achieve optimal results.

Here only the CPU dimension is considered for

placement. This method reduces number of active

PMs, overloaded PMs and number of migrations. It

outperforms 25-60% performance over traditional

non-multi-core placement.

This paper[4] initial VM placement is done in an

unloaded datacenter to reduce resource

wastage(CPU,Memory,BW) , power consumption

and to minimize SLA violation(MIPS).ACO (Ant

Colony Optimization) is used over GA and

Heuristic method.Genetic Algorithm doesn’t use

feedback information, searches with blindness

,large search space, do a lot of redundant iteration,

convergence speed and efficiency of optimum

solution is slow and low.Heuristic Method:Single

point search, fall into local optimum search, can’t

get global optimal solution, may not produce

any solution .ACO : Positive feedback

mechanism, pheromone is constantly updated,

optimal solution by efficient convergence .Findings

:Reduces resource wastage, power

consumption and SLA

violation

Objective[5] : To increase resource

utilization, to meet SLA requirement and to

reduce no. of PMs Method : Pearson

correlation Coefficient --Correlation between a pair

of VMs is considered Result : CPU

utilization increased, SLA violation

decreases(capacity requirement), number of active

servers reduced Issues : only

CPU demand is considered for VM multiplexing.

Multi-dimensional resources like(CPU,

Mem,Storage etc..) are not considered.

[6]mappingVMs to PMs is called VMP.VMP is

part of VM Migration.Goal is to minimize Energy

by shutting down services.4 Steps in VM

Migration:

a) Select PM which is

overloaded/underloaded

b) Select one or more VM

c) Select PM where selected VM can be

placed

d) Migrate VM to PM

The main goal of VM placement decision is

• Power Conservation

• SLA

• Revenue maximization

• Reliability

[7] designs an online VMP algorithm to increase

cloud providers revenue for the multi dimensional

resources. To develop an energy saving technique

using VM consolidation by migrating VMs to a few

active servers.

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com

Page 9 of 12

Page 5: A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

[8] This paper enhances the reliability of a cloud

service by increasing the fault tolerance level.

Replication based k-fault tolerance metric is used as

a backup when the primary PM fails. The prpsed

technique consist of three steps.One is host server

selection,second is optimal redundant VM

placement, and third one is recovery strategy

decision. the proposed approach consumes less

network resources.

[9] The objective of Decrease and Conquer

Genetic Algorithm (DCGA) is to find a new

optimal VM placement plan such that the total

energy consumption by all the servers, the total

number of VM migrations, and the computation

time can be minimized. DCGA outperforms FFD in

terms of energy consumption. DCGa minimizes

computation time and number of VM migrations

compared to classical GA.

VI. Classification of vm placement algorithms

VM placement algorithms are broadly classified in

to Power Based and Application QoS Based as

shown in figure 1.

Power based approaches aims at reducing the power

consumption in computing and cooling resources of

a data center. The vm to pm mapping is performed

in such a way to minimize the power usage by

shutting down unused pms.

Application based approaches aims at maximizing

the quality of service of a datacenter. These

approaches maximizes the resource utilization and

also minimizes the cost .

Classification of VM placement algorithms are

Constraint programming, Bin packing problem,

stochastic integer programming, Genetic algorithm,

Adaptive algorithms.

Heuristics to select PM

• First Fit

• Least Full First

• Most Full First

• Next Fit

• Random

• Tag & Pack

• Single Dimensional Best Fit

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com

Page 10 of 12

Page 6: A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

• Dot Product Based Fit

Heuristics to choose Vm

• Single Threshold

• Minimization of Migrations

• Highest Potential Growth

• Random Choice

The environment scenario for the VMP is

• Static/ dynamic

• Multiple resource types/ single type

• Fixed price/auction based

• improve csp revenue/ improve cu revenue

Algorithm

• Generate tasks

• Create VM list

• Submit task and VM to broker

• Create data center

• Create Host machines

• For all the available VMs

– Choose VM to assign to a first

available Host

– Assign VM to Host

• Send tasks to VMs

VII. REFERENCES

[1] K.Mills,J.Filliben and

C.Dabrowski,”Comparing VM-Placement

Algorithms for On-Demand Clouds”, Third

IEEE International Conference on Cloud

Computing Technology and Science, 2011

[2] Anton Beloglazov and Rajkumar Buyya,”

Energy Efficient allocation of Virtual machines

in Cloud Data Centers”, 10th IEEE/ACM

International Conference on Cluster, Cloud and

Grid Computing, 2010.

[3] Zoltan Adam Mann, Multicore-aware virtual

machine placement in cloud data centers , IEEE

Transactions on Computers,2015

[4] Fei MA,Feng LIU,Zhen LIU, Multi-objective

Optimization for Initial Virtual Machine

Placement in cloud Data Center, Journal of

Information & Computational science 9: 16

(2012) 5029-5038

[5] Zar Lwin Phyo and Thander Thein, Correlation

Based VMs Placement Resource Provision,

International Journal of Computer science &

Information Technology(IJCSIT) vol 5, no 1 ,

February 2013

[6] Rajeev Kumar Gupta and R.K.Pateriya, “

Survey on Virtual machine Placement

Techniques in Cloud Computing Environment”,

International journal on Cloud Computing:

Services and Architecture(IJCCSA), Vol 4,

No.4,August 2014

[7] Laiping Zhao,Liangfu,Zhou Jin, and Ce Yu,”

Online Virtual machine placement for

Increasing Cloud Providers Revenue “ IEEE

Transactions Services Computing,2015

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com

Page 11 of 12

Page 7: A Technical Review on Cloudsim based VM Scheduling ...ripublication.com/ijaerspl2019/ijaerv14n5spl_02.pdf · A Technical Review on . Cloudsim based VM Scheduling Techniques in . Cloud

[8] Ao Zhou, Shangguang Wang, , Bo Cheng, Zibin

Zheng, r, IEEE, Fangchun Yang, , Rong N.

Chang, Michael R. Lyu, and Rajkumar Buyya,

Cloud Service Reliability Enhancement via

Virtual Machine Placement Optimization, IEEE

TRANSACTIONS ON SERVICE

COMPUTING, VOL. XX, NO. XX, X XXXX

[9] Chanipa Sonklin, Maolin Tang, Yu-Chu Tian, A

decrease-and conquer genetic algorithm for

energy efficient virtual machine placement in

data centers,IEEE 15th International Conference

on Industrial Informatics(INDIN),2017

International Journal of Applied Engineering Research ISSN 0973-4562 Volume 14, Number 5, 2019 (Special Issue) © Research India Publications. http://www.ripublication.com

Page 12 of 12