An Empirical Study on the Impacts of Autonomy of Components on Qualities of Software Systems

More and more autonomous computing entities are implemented and deployed on the Internet and they are supposed to be able to adapt to the instable connection, decentralized control, dynamism and openness of the networking environment. Prior to implementing these autonomous computing entities, software engineers should decide a range of acceptable autonomy of components to ensure qualities of both individual components and the whole system. In order to qualitatively investigate how the autonomy of components impact on the qualities of the whole system and what other factors impact their relations, we conducted an experimental study based on the stochastic process. First, we give the definition of autonomy and an approach for measuring autonomy degree of a component based on the general recognition of the academia. Next, we build up a mathematical model for the relationship between autonomy degree and quality by using the stochastic process. At last, we construct an intelligent traffic control simulation system composed of Autonomous Components to concrete the mathematical model and to draw some generic conclusions from the experimental system. By recording these qualities under different autonomy degrees and different environment complexities, we work out the probability density distribution of quality movement. Combining the mathematical model, we give out some guides for autonomous components to adjust their autonomy degrees automatically under different contexts.