MODELING THE PERFORMANCE OF COMMUNICATION SCHEMES ON NETWORK TOPOLOGIES

JAN LEMEIRE, ERIK DIRKX, WALTER COLITTI
2008 Parallel Processing Letters  
This paper investigates the influence of the interconnection network topology of a parallel system on the delivery time of an ensemble of messages, called the communication scheme. More specifically, we focus on the impact on the performance of structure in network topology and communication scheme. We introduce causal structure learning algorithms for the modeling of the communication time. The experimental data, from which the models are learned automatically, is retrieved from simulations.
more » ... e qualitative models provide insight about which and how variables influence the communication performance. Next, a generic property is defined which characterizes the performance of individual communication schemes and network topologies. The property allows the accurate quantitative prediction of the runtime of random communication on random topologies. However, when either communication scheme or network topology exhibit regularities the prediction can become very inaccurate. The causal models can also differ qualitatively and quantitatively. Each combination of communication scheme regularity type, e.g. a one-to-all broadcast, and network topology regularity type, e.g. torus, possibly results in a different model which is based on different characteristics. Fig. 2. Network Topologies and Communication Schemes can be characterized by generic performance properties (a) or every combination entails a specific model (b). Next, we investigated the existence of a generic performance prediction model. The question is under which circumstances and to what extent can communication schemes and network topologies be characterized by generic performance properties. Such that the communication time, T comm , can be predicted by a simple model based on these characteristics, as shown in Fig. 2(a) . Each NT i is characterized by a Performance Factor, called
doi:10.1142/s012962640800334x fatcat:cn2gwemqe5covfsc7jk6escg3i