Logical cluster construction in a grid environment for similar protein retrieval

Abstract

This paper proposes a framework of constructing appropriate logical clusters in a grid environment for developing the protein surface retrieval system, which is a useful computational technology for protein functional analysis. In this system, the retrieval is processed with two layers of parallel processing by dividing computing resources into logical clusters. The agency host submits the job of calculating similarity between an input protein and each active site in the database to every logical cluster in the Grid. Each logical cluster calculates the similarity using the MPI-based parallel program and returns the result to the agency host. We also present a method of suggesting a better structure pattern of logical clusters based on the estimation of the processing time for surface comparison in each logical cluster. The experimental results prove the effectiveness of the proposed framework.