Deployment method of virtual machine cluster based
on energy minimization and graph cuts theory*
Zhiping Peng, Bo Xu*, Delong Cui
Department of Computer Science and
Technology, Guangdong University of
Petrochemical Technology
Maominig, China
E-mail: xubo807127940@163.com
Weiwei Lin
School of Computer Science and
Engineering, South China University of
Technology
Guangzhou, China
XuAn Wang
Engineering University of Chinese
Armed Police Force
Xi'an, China
Abstract—A deployment method had been constructed for
virtual machine cluster based on energy minimization and graph
cut theory. First, virtual machine cluster was described by energy
minimization. Second, deployment of virtual machine cluster was
changed into maximum flow minimum cut problem. Thirdly,
sink and source point was added and activity node was found.
Finally, activity tail-nodes were connected and cut formed for
virtual machine cluster. Experimental results show that the
proposed method can effectively cut virtual machine cluster and
select rationally physical host for sub-cluster.
Keywords—virtual machine cluster
deployment of virtual
machine
maximum flow and minimum cut
energy
minimization
I. INTRODUCTION
In various supporting technologies for cloud computing,
virtualization and virtual machine deployment are critical [1].
When resources are required to complete user tasks, they are
directly mapped to hardware resources, which usually entail
problems such as low efficiency and inaccurate estimation of a
physical host’s current status [2]. Therefore, the task
requirements in cloud computing typically need to be translated
by virtualization technology, which is initially deployed in a
virtual machine and subsequently deployed in an actual
physical host.
Virtualization technology significantly reduces the level of
hardware coupling for task requirements and improves the
service efficiency of cloud computing, i.e., how to deploy a
virtual machine in a physical host or a virtual machine
deployment problem [3]. Chen et al. employed the disk’s
Copy-on-Write mode as a starting point, divided the disk
mapping file into several smaller parts and subsequently
implemented virtual machine deployment. This method may
improve resource utilization and enable the attainment of a
higher speed [4]. Wen et al. noted that the majority of virtual
machine deployment algorithms employ central processing unit
(CPU) utilization as their primary hardware index; however,
this strategy is likely to cause load imbalance. To overcome
this limitation, they considered the resource consumption of a
virtual machine in the CPU forecast. Thus, an automatic
selection mechanism based on the virtual machine deployment
method was developed [5]. Li et al. proposed that accelerating
a virtual machine’s deployment speed is a critical aspect of
cloud computing. He developed a distributed mirror storage
based deployment method, which can reduce the complexity of
the deployment process and improve a virtual machine’s
deployment speed [6]. Feng et al. employed the Infrastructure
as a Service (IaaS) mode cloud as a case study to investigate
the strategy of deploying clustered virtual machines in
clustered physical hosts. During deployment, he quantified and
sorted all available virtual machine resources and physical
hosts and extended the descending adaptive optimal method to
obtain a configuration between the two sets. Compared with
the pure virtual machine deployment method, this method
exhibits certain improvements in total efficiency [7]. Yang et
al. proposed that the virtual machine deployment problem has
gradually evolved from single virtual machine deployment to
bulk virtual machine deployment. They developed an ant group
optimize algorithm to address the bulk virtual machine
deployment problem and obtained positive deployment results
[8]. Wang et al. suggested that a series of related virtual
machines that are deployed in the form of clusters are more
likely to achieve total efficiency of cloud computing and cloud
service. Based on this idea, he separately grouped virtual
machine resources and physical hosts into clusters and matched
the clusters to the deployed virtual machine clusters [9].
II. V
IRTUAL MACHINE DEPLOYMENT METHOD BASED ON
ENERGY MINIMIZATION GRAPH CUT THEORY
A. Energy minimization problem in virtual machine cluster
segmentation deployment
In the early stage of virtualization technology, a user
request is normally configured into a series of virtual machine
resources. When these virtual machine resources are
configured in hardware, the connections between virtual
machines are typically not considered; the only consideration
is which physical host is more suitable for deployment of
which virtual machines. When connections between related
virtual machines are examined after physical deployment, they
will be constrained by the physical host that hosts the virtual
machines [10]. Therefore, the main focus of the current
research is to group virtual machines that are related to a
certain user’s task service into a cluster and configure physical
2015 10th International Conference on P2P, Parallel, Grid, Cloud and Internet Computing
978-1-4673-9473-4/15 $31.00 © 2015 IEEE
DOI 10.1109/3PGCIC.2015.51
800
2015 10th International Conference on P2P, Parallel, Grid, Cloud and Internet Computing
978-1-4673-9473-4/15 $31.00 © 2015 IEEE
DOI 10.1109/3PGCIC.2015.51
800