Abstract
With the rapid development of the Web, cloud computing has received much attention. The key enabling factor for cloud computing is the virtualization technology by which multiple virtual machines can run in a shared physical machine and enable high utilization of hardware resources. However, the consolidation also introduces contention in shared resources, leading to degraded application performance. Then, one of challenges is to detect and mitigate the performance crisis where one or more virtualized machines’ performance target is not achieved. We present a method for mitigating the performance crisis, which leverages the performance interference model to predict the possibility of the performance crisis caused by the consolidation and employs this model to learn the bad or good consolidation plan to identify a new placement strategy to mitigate the performance crisis. Our experiments show the better performance of the mitigating methods.
Acknowledgement
The authors would like to thank the reviewers for their detailed reviews and constructive comments, which have helped improve the quality of this paper.