Evolutionary Algorithm Performance Profiles on the Adaptive Distributed Database Management Problem

Evolutionary algorithms have been shown to be effective in providing configuration optimisation to dynamic load balancing in distributed database systems and Web servers. This paper explores the tuning parameter performance profile of such techniques over a variety of problems, including the adaptive distributed database management problem (ADDMP), focusing on a range of interesting and important features. The ability of the evolutionary search process to reliably find good solutions to a dynamic problem in a minimal and consistent run-time is of paramount importance when considering their application to real-time industrial control problems. This paper demonstrates the existence of certain optimal parameter values, particularly for the rate of applied mutation, which are shown to produce consistently good problem solutions in a low number of evaluations with a minimum standard deviation. [PUBLICATION ABSTRACT]