An energy-efficient VM placement method for cloud data centers using a hybrid genetic algorithm
| Date | 12 November 2018 |
| Published date | 12 November 2018 |
| Pages | 430-445 |
| DOI | https://doi.org/10.1108/JSIT-10-2017-0089 |
| Author | Mohamed Amine Kaaouache,Sadok Bouamama |
| Subject Matter | Information & knowledge management,Information systems,Information & communications technology |
An energy-efficient VM placement
method for cloud data centers
using a hybrid genetic algorithm
Mohamed Amine Kaaouache and Sadok Bouamama
University of Manouba, Ecole Nationale des Sciences de l’Informatique,
Manouba, Tunisia
Abstract
Purpose –This purpose of this paper is to propose a novel hybrid genetic algorithm based on a virtual
machine (VM) placement method to improve energy efficiencyin cloud data centers. How to place VMs on
physical machines (PMs) toimprove resource utilization and reduce energy consumptionis one of the major
concerns for cloudproviders. Over the past few years, many approaches for VM placement(VMP) have been
proposed; however, existing VM placementapproaches only consider energy consumption by PMs, and do
not considerthe energy consumption of the communication networkof a data center.
Design/methodology/approach –This paper attemptsto solve the energy consumption problem using
a VM placement method in cloud data centers. Ourapproach uses a repairing procedure based on a best-fit
decreasing heuristic to resolve violations caused by infeasible solutions that exceed the capacity of the
resourcesduring the evolution process.
Findings –In addition, by reducingthe energy consumption time with the proposed technique,the number
of VM migrations was reduced comparedwith existing techniques. Moreover, the communication network
caused lessservice level agreement violations (SLAV).
Originality/value –The proposed algorithm aims to minimize energy consumption in both PMs and
communication networks of data centers.Our hybrid genetic algorithm is scalable because the computation
time increasesnearly linearly when the number of VMs increases.
Keywords Energy, Cloud computing, Hybrid genetic algorithm, Server consolidation,
Virtual machine placement
Paper type Research paper
1. Introduction
Cloud computing has recently evolved as one of the most promising technologies in the
current information technology (IT) scenario and promises virtually unlimited resources
(e.g. networks, servers, storage, applications and services) that can be rapidly provisioned
and released with minimal management effort. As a new style of computing in the
ubiquitous computing paradigm that offers IT resources through the internet by solving
complex IT-related problems at a lower cost and in less time, cloud computing faces some
new challenges. One of the prominentissues is the energy efficiency of data centers.
As stated by Amazon’s estimations, the energy-related costs at its data centers account
for 42 per cent of the total operatingcost, inclusive of both direct power consumption (19per
cent) and investment in the supportinginfrastructure for cooling and power distribution (23
per cent). Yet, theaverage data center energy efficiency is merely 50 per cent. In addition,the
ever increasing energy consumption leads to a considerable increase in carbon dioxide
emissions. Thus, reducingenergy consumed by data centers is considered a major issue.
According to recent studies, data centers will consume 2 per cent of total global energy
consumption by 2020. The demand for overall energy requirements at data centers is rapidly
JSIT
20,4
430
Received8 October 2017
Revised13 December 2017
17September 2018
29September 2018
Accepted15 October 2018
Journalof Systems and
InformationTechnology
Vol.20 No. 4, 2018
pp. 430-445
© Emerald Publishing Limited
1328-7265
DOI 10.1108/JSIT-10-2017-0089
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1328-7265.htm
increasing at the rate of 18 per cent every year. For example, the projection of total power
consumed by US data centers, which is now at 70 billion kWh/year, will be 73 billion kWh/year.
Emerson Network Power’s analysis shows how energy is used within a data center
(Power, 2009). It was found thatenergy consumed by servers’accounts for 52 per cent of the
total consumption, whereas support systems consume the remaining 48 per cent.
Furthermore, everywatt of savings that could be achieved on servers created approximately
2.84 W of savings overall. Hence, it becomes critical to determine an effective way to
improve the energy efficiencyof servers.
To improve the energy efficiency of data centers, server consolidation using
virtualization technology has become an important technology that solves this problem
(Clark et al.,2005). The basic idea behind server consolidation technology is to migrate
virtual machines (VMs) to as few energy-efficientphysical machines (PMs) as possible, and
then switch off or hibernate all the other PMs especiallyduring off-peak traffic hours. There
are many solutions listed in the literature, which basically can be divided into two
categories: designing energy-proportional servers or networks and establishing energy-
aware virtualization over servers (Usmani and Singh, 2015). The last approach has the
purpose of improvingthe energy efficiency of servers.
This problem of server consolidation is basically a VM placement problem. In the past
few years, many approaches for the VM placement problem have been proposed (Verma
et al., 2008). However, existing VM placement approaches do not consider the energy
consumption in the communication networks of data centers. Energy consumption in
communication networks is not trivial and, therefore, should be considered in the VM
placement problemto make data centers more energy efficient.
The VM placement problem can be regarded as a bin-packing problem by considering
PMs as bins, and the VMs to be placed can be considered as objects to be filled in the bin,
and this has been confirmed to be an NPC (non-deterministicpolynomial complete) problem
(Beloglazov and Buyya, 2012). To lower operating costs by saving energy, the concept of a
green cloud has been proposed. A genetic algorithmhas been proved to be a solution for the
VM placement problembecause of its advantages of speediness and adaptability.
To further improve the efficiency and performanceof data centers, this paper presents a
Hybrid Genetic Best Fit decreasing algorithm for Bin Packing (HGBF_BP) in solving the
energy-efficient VM placement problem. Experimental results show that HGBF_BP
significantly outperforms the other policies and the original Genetic Algorithm (GA) and
that the proposed approachscales up well when the problem size increases.
The remaining paper is structured as follows. In Section 2, we present related work. In
Section 3, we formulate the VM placement problem. Section 4 details the HGBF_BP
algorithm and describes its adaptation for VM placement. Section 5 evaluates the
performance, efficiency and scalability of HGBF_BP, and finally, in Section 6, we make
conclusions and discuss this work. We also talk about our future work on the energy-
efficient VM problem.
2. Related work
Energy consumption is an importantissue in many fields of research. Both consumers and
industries want their productsto use less power, to reduce energy costs. As systems become
larger and more complex, theytypically consume more energy. This problem is extendedto
networks, and it consequently extends to cloud computing. Because data centers contain
large clusters of computers, any reduction in energy expenditure can result in large
economic savings. For this reason, there has been much research performed on how to
reduce energy consumptionand on how to reduce energy consumption within a network.
An energy-
efficient VM
placement
method
431
To continue reading
Request your trial