An evaluation of globus and legion software environments
ICCS'03 Proceedings of the 2003 international conference on Computational science: PartII
| Hi-index | 0.00 |
In this article our goal is to apply the mobile agent technology to provide a better scheduling to MPI applications executing in a cluster configuration. This approach could represent in a distributed cluster environment an enhancement on the load balancing of the parallel processes. MPI in cluster of heterogeneous machines could lead parallel programmers to obtain frustrated results, mainly because of the lack of an even distribution of the workload in the cluster. As a result, before submitting a MPI application to a cluster, we use our JOTA mobile agent approach to acquire a more precise information of machine麓sworkload. Therefore, with a more precise knowledge of the load and characteristics in each machine,we are ready to gather lightweight workstations to form a cluster. Our empirical results indicates that it is possible to spend less elapsed time when considering the execution of a parallel application using the agent approach in comparison to an ordinary MPI environment.