Performance modeling based on artificial neural network in virtualized environments

ISCAS OpenIR

	Performance modeling based on artificial neural network in virtualized environments
	Du, Guangyu (1); He, Hong (1); Meng, Fanxin (1)
	2013
发表期刊	Sensors and Transducers
ISSN	17265479
卷号	153 期号:6 页码:37-44
摘要	Large-scale data centers leverage virtualization technology to achieve excellent resource utilization, scalability and high availability. Although virtualization technology has the advantages such as fault isolation, environmental isolation and security isolation, current virtualization techniques do not have effective performance isolation among virtual machines. The hidden resource competition does exist which is especially severe for applications running on the same physical machine. It is essential to build models to accurately predict application performance interference among virtual machines to mitigate performance interference effect. In this paper, we mainly focus on performance modeling in virtualized environments. We explore modeling techniques of artificial neural network and regression models and evaluate their effectiveness in modeling application performance in virtualized environments. Based on the performance prediction model, we propose the resource management architecture. Experimental evaluations show that our performance model has good prediction performance over regression models. © 2013 IFSA.; Large-scale data centers leverage virtualization technology to achieve excellent resource utilization, scalability and high availability. Although virtualization technology has the advantages such as fault isolation, environmental isolation and security isolation, current virtualization techniques do not have effective performance isolation among virtual machines. The hidden resource competition does exist which is especially severe for applications running on the same physical machine. It is essential to build models to accurately predict application performance interference among virtual machines to mitigate performance interference effect. In this paper, we mainly focus on performance modeling in virtualized environments. We explore modeling techniques of artificial neural network and regression models and evaluate their effectiveness in modeling application performance in virtualized environments. Based on the performance prediction model, we propose the resource management architecture. Experimental evaluations show that our performance model has good prediction performance over regression models. © 2013 IFSA.
收录类别	EI
部门归属	(1) School of Mechanical Electric and Information Engineering, Shandong University, Weihai, 264200, China; (2) Institute of Software of Chinese Academy of Sciences, Beijing, 100190, China
语种	英语
内容类型	期刊论文
URI标识	http://ir.iscas.ac.cn/handle/311060/17046
专题	中国科学院软件研究所
推荐引用方式 GB/T 7714	Du, Guangyu ,He, Hong ,Meng, Fanxin . Performance modeling based on artificial neural network in virtualized environments[J]. Sensors and Transducers,2013,153(6):37-44.
APA	Du, Guangyu ,He, Hong ,&Meng, Fanxin .(2013).Performance modeling based on artificial neural network in virtualized environments.Sensors and Transducers,153(6),37-44.
MLA	Du, Guangyu ,et al."Performance modeling based on artificial neural network in virtualized environments".Sensors and Transducers 153.6(2013):37-44.