Generating Adaptation Rules of Software Systems
Yang Liu, Di Bai, Wenpin Jiao · 2018
Nowadays, applications are deployed and executed in open, uncertain and dynamic environments. Increasingly, applications are required to have the capability to automatically change its behaviors in response to changing environmental conditions. As the complexity of adaptive and autonomic systems grows, designing and managing the set of adaptation rules becomes increasingly challenging and may produce huge computation cost. If we dynamically generate adaptation rules at run time, it's difficult to deal with the changes quickly. The software system challenges the method for efficiently generating effective adaptation rules. This paper proposes an approach to combine genetic algorithm and linear regression to automatically generate adaptation rules for software systems. Unlike traditional rule-based adaptation methods, our solution enables a system to obtain a prediction function to determine the corresponding system configuration under any considered environmental condition. We have applied this genetic algorithm based approach to the dynamic reconfiguration of two different software systems. The experimental results show that our method is practical and highly-efficient in software reconfiguration under changing environmental conditions. Besides, the user's requirements can be well satisfied.