Abstract

Due to the increasing interest to achieve high performance computing using cluster, which consists of heterogeneous workstations and/or personal computers (PCs) connected via a fast network. Usually programming such systems is done by applying message-passing libraries like MPI. Due to the lack of a load balancing facility, it is quite possible to overload machines on the network with tasks, and lead parallel applications to frustrated results. In this paper, the mobile agent based approach to improve the load balancing of parallel tasks in heterogeneous computing environment has been proposed. Its benefit has been demonstrated with the experimental results.

Key words: Cluster of heterogeneous workstations, mobile agent, load balance, parallel computing.