Performance Estimation of a BOINC-Based Desktop Grid for Large-Scale Molecular Docking

Abstract

This paper addresses the performance evaluation of a heterogeneous distributed computing environment (Desktop Grid) for largescale molecular docking and scoring. At the initial stage of a distributed computing project based on a Desktop Grid, the set of computational nodes is dynamically growing. With peculiarities of Desktop Grids such as heterogeneity and unreliability of the nodes, this factor imposes difficulties on the task scheduling and algorithm scaling. We analyze a Desktop Grid performance, propose the efficiency metrics for this stage and provide statistics and analysis of the initial stage of a volunteer computing project named SiDock@home that is performing large-scale in silico medicinal chemistry experiments.