Performance Testing for IP Services and Systems

Abstract

The dynamic world of IP services and systems is focused on rapid time-to-market in the face of high demand uncertainty. Performance testing is frequently either omitted or performed in a cursory manner. Successful IP services experience phenomenal growth, and are often unprepared for the performance problems that they encounter while scrambling to scale to meet rapid demand increases. This growth spurt poses particular challenges for performance analysts. In this paper we describe methods for assessing performance through empirical performance testing combined with “gray-box” modeling. The paper defines the performance testing process and shows through a case study how the behavior observed in the stress tests can provide developers with opportunity areas for performance improvement. Based on the success of this approach, we recommend that software stress testing be incorporated formally into the software development process, especially when the development consists largely of integrating external vendor components.