An empirical study into component system evolution

Abstract

When evolving component based systems, possible side effects can bemitigated by changing only what is necessary. Identifying this minimal change is non-trivial, and exploring different heuristics used for this is the focus of this paper. We look at three different heuristics and compare their properties while simulating evolution of an Ubuntu GNU/Linuxdistribution. This simulation involves randomly selecting components to install, then calculating the resulting system with respect to aheuristic over many generations. We look at the mean and volatility of change to a system to compare different heuristics, and find that simple heuristics can result in minimal changes just as well as other more complex heuristics.