Modelling communication performances in heterogeneous and hierarchical environments

Abstract

Today, due to many reasons, such as the inherent heterogeneity, the diversity and the continuous evolution of actual computational supports, writing efficient parallel applications on such systems represents a great challenge. To achieve good performances in these execution environments, the impact of communications on the global performances of parallel applications should not be neglected. Our objective within this work is to design an analytical model able to predict accurately the performance of communication operations on computational systems characterised by both heterogeneity and hierarchical structure. We principally aim to guarantee the accuracy of the predicted performances with a neglected additional overhead. The proposed model was applied on point-to-point and collective communication operations and showed by achieving experiments on a hierarchical cluster-based system with heterogeneous resources that the predicted performances are close to measured ones.

Keywords:

• performance modelling and prediction
• communication models
• hierarchical clusters
• parallel applications

Notes

^1. http://www.top500.org.