Large-scale engineering simulation typically requires resources beyond that of a single workstation. On the other hand, given a large number of workstations, servers and personal computers, due to the communication cost involved, it may make no sense to use all of the available resources to run a single simulation. In fact, in such a situation, it makes better sense to run multiple simulations, each over its own cluster of workstations. In this paper we have tried to find optimal clustering of a set of highly heterogeneous computing resources,considering similarity of machine availability pattern plus the interconnection speed, in the context of a campus grid. In this work, we have studied application of Genetic Algorithm with and without seeding towards automatically tuning FDTD performance. Experimental result suggests that, the specific seeding rate, tune the GA automatically for optimal resource clustering of FDTD compuation. © 2012 Elsevier B.V.