Optimization of Computational Efficiency of GP2U-Accelerated Grid Applications on Non-dedicated Desktop Grids: A Speed-Up Factor Analysis

Phenomenal growth in the field of biology has motivated computer science to scale up high-performance computing (HPC) frameworks to cater the ever-growing demand of today’s biologists. High cost of HPC hinders the progress of study in financially weak research teams, so non-dedicated desktop grid is makeshift arrangement for such research groups, in which there is a set of existing desktop computers. Gaining maximum possible performance out of such non-dedicated desktop grid is a challenge. In this paper, an attempt has been made to gain maximum speed-up of grid-enabled application deployed in a non-dedicated desktop grid. The grid-enabled application developed has been GPU accelerated, and load balancing algorithms’ performance parameters have been optimized along with node provisioning resulting an overall speed-up of 16.01 with eight nodes only.