Binary Information Press
Scalability is an important concept in the domain of parallel computing. Since Graphics Processing Unit (GPU) clusters are and will be widely utilized in high performance computing platforms, the authors investigate the factors influencing the scalability for combinations of Parallel Algorithms (PA) and GPU Clusters (GC). They present a scalability model for combination PA-GC and then validate it by executing some well-known benchmarks on certain GPU cluster (i.e., TH-1A). Experimental results show that the proposed model can well predict the scalability of various PA-GC combinations.