Workload characterization of modern computational biology applications

As computational biology emerges as an important class of scientific computing applications, it is becoming more evident that further advances in the field will render such applications even more important for pharmaceutical research, protein structure prediction and development of gene-based therapies. We expect that the performance of computational biology applications will therefore become an important factor in denning future high, performance computing systems. This paper presents a study of the execution characteristics of several computational biology applications in an effort to understand the general traits of such workloads. A performance analysis of a set of benchmarks chosen to represent major classes of commonly used computational biology applications is presented, and the results briefly discussed.