A KCCA-based Approach for Performance Modeling of Web Application in Clouds
Haiyan Wang, Zhenzhong Zhang, Bin Zhao, Limin Xiao, Li Ruan · 2013
Cloud computing which services in a pay-as-you-go manner to the users transparently, is a newly execution environment. However, the complex and dynamic relationships between application workload and performance and resource consumptions make it difficult to manage the resources in the cloud. Performance modeling provides the ability to predict the performance of application and resource consumptions. In this paper, we describe a dimensionality reduction algorithm called Kernel Canonical Correlation Analysis (KCCA) to predict the performance and resource consumptions of web application in the cloud computing environment. This approach provides much more expressiveness in capturing similarity and its correlations can then be used to quantify performance similarity. With KCCA, we build a single model for multiple performance metrics and reduce the dimensionality of both the workload and performance. Then we detail the effectiveness of the approach using example data gathered from production environment. The performance evaluation results demonstrate that our methodology could effectively predict the performance of web applications.