Performance analysis of service-oriented architectures with multi-factor sensitivity analysis

This paper presents a continuous investigation on performance analysis with sensitivity analysis. As an effort to develop a systematic approach to improve the performance of service-oriented software systems, the original work introduced a statistical approach of two-factor-based sensitivity analysis to software performance analysis. The goal of generating accurate performance feedback, however, was only partially achieved as performance analysis needs to consider more factors. This paper presents a generalization for the statistical method to handle multiple factors. In addition, it gives detailed discussions on sensitivity analysis with three factors, and provides experiment results to demonstrate the need and advantages of analyzing multiple factors at the same time.