International Journal of Software Engineering and Its Applications Vol. 4, No. 1, January 2010 Availability-Based Software Performability Model with User-Perceived Performance Degradation Koichi Tokuno and Shigeru Yamada Department of Social Management Engineering, Graduate School of Engineering, Tottori University, Japan
[email protected],
[email protected] Abstract This paper discusses software performability evaluation considering the real-time property. We assume that the software system has two operational states from the viewpoint of the end users: one is operating with the desirable performance level according to specification and the other is with degraded performance level. The time-dependent behavior of the system is described by the Markovian software availability model with performance degradation. Assuming that the system can process the multiple tasks simultaneously, we analyze the distribution of the number of tasks whose processes can be completed within the processing time limit with the infinite server queueing model. We derive several software performability measures; these are given as the functions of time and the number of debugging activities. Finally, we illustrate several numerical examples of the measures to investigate the impact of consideration of the performance degradation on the system performability evaluation. Keywords: performability, real-time property, performance degradation, software availability model, infiniteserver queueing model 1. Introduction The studies on performability evaluation methods for hardware-oriented computing systems have much been discussed [1,2,3,4]. However, on the other hand, most of studies on software-oriented reliability evaluation have treated only the inherent reliability characteristics such as the residual fault content, the mean time between software failures (MTBSF), and the software reliability function [5,6]. However, recently, software-conscious approaches extended to performability evaluation have also increased [7,8,9]. Most of the existing software-conscious approaches are discussed on the basis of performability measures in steady states and assume that the probabilistic or stochastic characteristics in system failure and restoration do not change even though the system is debugged or refreshed, i.e., the system returns to the initial condition in terms of the failure and restoration characteristics, neither better nor worse states. As to this point, the analytical framework in the above studies is basically similar to the hardwar