SLA-oriented Research on Prediction and Evaluation of Service Components Performance

The application services which deployed in cloud environment have some characteristics, such as the variability of concurrent requests and the differences of resource demands among components, which bring certain influence on service performance even cause some potential problems. Therefore, making an effective service performance prediction mechanism has become a research hotspot for application services based cloud service components. This paper proposes SLA-oriented research on prediction of service component and introduces the fundamental process of performance prediction. It adopts the non-negative matrix factorization method for performance prediction and makes an evaluation on component performance. The experiment result indicates the method in this paper is feasible and valid, and the accuracy is higher than the traditional approaches.