The Journal of Supercomputing, 26, 77–94, 2003 # 2003 Kluwer Academic Publishers. Manufactured in The Netherlands. Communication Delay in Wormhole-Switched Tori Networks under Bursty Workloads G. MIN Department of Computing, University of Bradford, Bradford, BD7 1DP, UK M. OULD-KHAOUA Department of Computing Science, University of Glasgow, Glasgow, G12 8RZ, UK Abstract. Workloads generated by the real-world parallel applications that are executed on a multicomputer have a strong effect on the performance of its interconnection network—the hardware fabric supporting communication among individual processors. Existing multicomputer networks have been primarily designed and analysed under the assumption that the workload follows the non-bursty Poisson arrival process. As a step towards obtaining a clear understanding of network performance under various workloads, this paper presents a new analytical model for computing message latency in wormhole switched torus networks in the presence of bursty traffic, based on the well-known Markov-Modulated Poisson Process (MMPP). In order to derive the model, the approach for accurately capturing the properties of the composite MMPPs is applied to characterize traffic on network channels. Moreover, a general method has been proposed for calculating the probability of virtual channel occupancy when the traffic on network channels follows a multi-state MMPP process. Simulation experiments reveal that the model exhibits a good degree of accuracy. Keywords: multicomputers, interconnection networks, multimedia applications, message latency, performance modeling/analysis 1. Introduction Over the past few years, we have witnessed an explosive growth in multimedia applications that combine different media types, e.g., audio, video and text, with different quality-of-service requirements. Example applications include WWW servers, video-on-demand servers, interactive simulation, video-conferencing and collaborative design environments [11]. The potential market of the emerging multimedia applications is large enough to even justify specific architectures oriented to support them efficiently. Multicomputers are widely accepted as good candidates for large-scale parallel multimedia servers as they can meet the high computation and communication requirements of multimedia applications, offering high- processing power, reliability, and above all, high-bandwidth I/O at low cost [11]. Representative examples are the Oracle Media Server running on the N-Cube parallel computer [5] and the MPEG video parallel encoder running on the Intel Paragon [25].