Resource onitoring and rediction in loud
omputing nvironments
Hanxiong Chen
School of Telecommunication and Information Engineering
Nanjing University of Posts and Telecommunications
Nanjing, China
562352499@qq.com
Xiong Fu, Zhongrui Tang, Xinxin Zhu
School of Computer Science and Technology
Nanjing University of Posts and Telecommunications
Nanjing, China
fux@njupt.edu.cn
Abstract—Cloud computing provides elastic, scalable
resource sharing service by resource management. Resource
monitoring and prediction are the foundation to achieve resource
automation, high performance management in cloud computing
environment. This paper addresses the resource monitoring and
prediction problem in cloud computing environment, designs and
implements an adaptive resource monitoring framework for
cloud computing, and presents a resource prediction mechanism
based on Vector Auto Regression (VAR) by the correlation
between various resources. Related experiments show that the
proposed resource monitoring framework can effectively monitor
the resource usage in cloud computing environment, and
prediction mechanism based on vector auto regression compared
to other prediction mechanism could be more effective to predict
resource usage.
Keywords—cloud computing; resource monitoring; resource
prediction; vector auto regression
I. INTRODUCTION
In recent years, with the rapid development of computing,
storage and communication technology, cloud computing,
which has been widely regarded as the third technological
revolution after personal computer and the Internet and will be
the key technology to lead industrial revolution over the next
20 years, has been proposed as a brand new commercial
network computing model for resources sharing[1,2].
According to the survey of a market research firm Gartner in
2010, cloud computing has been one of the technologies which
are most concerned about by the IT users, meanwhile many
large IT companies have established a large number of data
centers and provided external cloud computing resources
services, such as Google has deployed over 36 data centers and
computer node reaches millions, Microsoft's servers will be
doubled every 14 months, and only its cloud computing data
center servers located in Chicago is to reach hundreds of
thousands of units.
Although cloud computing has gained great development in
recent years, but still faces lots of the key technical problems,
the first is the resource management issues, particularly the
demand of resource management dynamic, flexible and
automated is more prominent[3]. Resource monitoring and
prediction in cloud computing environment is an important part
of resource management, is also the basis for other
management techniques of resource, Its role is mainly
manifested in the following aspects:(1)Through resource
monitoring and prediction, to understand resource use and task
performance, as the basis for resource allocation, task
scheduling and load balancing in cloud computing; (2)Through
resource monitoring and prediction, tracking resource
management, discovering and solving resource failure timely,
taking resource warning and alarm to maintain the stable
functioning of cloud computing; (3)Through resource
monitoring and prediction, to count resource usage of different
customers. Therefore, the study of resource monitoring and
prediction in the cloud computing environment is of great
significance to solve the management problems of cloud
computing and to promote the development of cloud
computing[4].
Though in the past people did much work for resource
monitoring and prediction in distributed computing and grid
computing, and developed appropriate monitoring systems,
such as DRMonitor[5], MDS[6], Ganglia[7], etc., but these
results were difficult to use in a cloud computing environment
for the following two reasons: on the one hand, in cloud
computing environment, the usage of a certain resource by user
is not only the basis for billing, but also the foundation of
resource scheduling or task migration. Existing monitoring
systems can not achieve such virtual machine level monitoring
and prediction. On the other hand, the data center in cloud
environment is large-scale. The resource consumption of large
quantities of computing, storage, network generated by original
resource monitoring and prediction is also likely to be a huge
burden of the cloud computing system which causes distress to
cloud computing system to provide resources services
efficiently.
In this paper, we address on the resource monitoring and
prediction problem in cloud computing environment, design
and implement a cloud-oriented adaptive resource monitoring
framework. The framework can not only implement virtual
machine level resource monitoring, meet the relevant
requirement of flexible resource usage billing, but also
adaptively adjust the frequency of monitoring by using the
resource prediction or system performance. Balance and reduce
resource consumption brought by resource monitoring in the
premise of meeting the need of resource management. In
resource prediction, the paper is based on the correlation
between all types of resources used by users when users rent
cloud computing services and present a vector auto regression
resource prediction mechanism. The prediction of the results
2015 3rd International Conference on Applied Computing and Information Technology/2nd International Conference
on Computational Science and Intelligence
978-1-4673-9642-4/15 $31.00 © 2015 IEEE
DOI 10.1109/ACIT-CSI.2015.58
288