Model-Driven Diagnostics of Underperforming Communicating Systems

Abstract

This paper proposes methods for improving the performance of a communicating system that has failed its performance test. The proposed methods extend our earlier published model-driven performance testing method, which automatically determines whether the tested system is able to serve the specified number of requests within a second in worst case while serving a specified number of users simultaneously. The underperformance diagnostic methods presented in this paper are given as an input the formal performance model representing the system under test, which was built up by our performance testing method in the performance testing phase. The presented methods aim at improving the performance of the system under test to the desired level at minimal cost. One of the methods presented in this paper is a binary linear program, while the other is a heuristic method which, according to our simulation results, performs efficiently.