An analytical model for hypercubes in the presence of multiple time-scale bursty traffic

摘要

The interconnection network is one of the most critical architectural components in a parallel computer as any interaction between individual processors ultimately depends on its effectiveness. Recent studies have shown that traffic workloads in many parallel computation environments reveal burstiness over a number of time-scales, which has a considerable impact on network performance. Analytical modelling plays an important role in obtaining a clear understanding of the performance of interconnection networks with various design spaces and operating under different traffic conditions. This paper proposes a concise analytical model for the hypercube operating under multiple time-scale bursty traffic. The validity of the model is demonstrated by comparing analytical results to those obtained through simulation experiments of the actual system. The analytical model is then applied to evaluate the impact of multiple time-scale bursty traffic on the performance of hypercubes and investigate the effects of Hurst parameters of the traffic. (C) 2006 Elsevier B.V. All rights reserved.