An adaptive load balancing algorithm using simple prediction mechanism

Abstract

In this paper, we present an adaptive location policy based on state polling using system state information according to a prediction mechanism. In each node, the system state information is efficiently collected and changed with low overheads at run time. Also, it is systematically maintained by using good data structures in each node. The prediction mechanism is employed to accurately decide the better destination node. We perform event-driven simulations and compare our algorithm with some of the well-known ones. We show that the results exhibit the relative performance improvement over other algorithms.