MVEI: An Interference Prediction Model for CPU-intensive Application in Cloud Environment

Consolidation and overbooking are two important methods to improve the resource utilization in cloud computing. In overbooking cloud environment, the resources occupied by virtual machine are restricted to guarantee performance fairness between multi-tenants. However, the interaction among co-host applications results in performance degradation because of the resource sharing and competition. Previous studies about performance interference mainly focused on traditional virtual environment and didn´t take the resource restriction into account. In order to predict the effect of interference in cloud environment, which can give suggestions for resource scheduling in turn, we conduct comprehensive experiments about CPU-intensive application in Cloud. Firstly, the interfering factors of the performance are analyzed, and the relationships between different factors and interference are obtained by using the least squares method. Then, a multivariable exponent interference (MVEI) model is proposed to predict the interference in resource restricted environment. Finally, the experimental results indicates the MVEI model can predict interference more precisely than linear and quadratic model.