Comparative studies of load balancing with control and optimization techniques

Load balancing is a widely used technique to optimizing distributed computing system performance. System response delays are reduced by equalizing the loads, such as adjusting memory pool sizes to balance disk access demands in a database management system. In this paper we formulate load balancing as a constrained optimization problem and investigate two load balancing controllers based on feedback control theory and optimization theory. We show the difference and equivalence between their design methods and criteria. Furthermore, our studies on a DB2 universal database server reveal their performance difference regarding to system noise and workload variations.