Publication Details
Overview
 
 
Jan Lemeire, Walter Colitti, Erik Dirkx
 

Contribution to journal

Abstract 

This paper investigates the influence of the interconnection network topology of a parallel system onthe delivery time of an ensemble of messages, called the communication scheme. More specifically, wefocus 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. Theexperimental data, from which the models are learned automatically, is retrieved from simulations. Thequalitative models provide insight about which and how variables influence the communication performance.Next, a generic property is defined which characterizes the performance of individual communicationschemes and network topologies. The property allows the accurate quantitative prediction ofthe runtime of random communication on random topologies. However, when either communicationscheme or network topology exhibit regularities the prediction can become very inaccurate. The causalmodels can also differ qualitatively and quantitatively. Each combination of communication schemeregularity type, e.g. a one-to-all broadcast, and network topology regularity type, e.g. torus, possiblyresults in a different model which is based on different characteristics.

Reference