Mapping and performance prediction for distributed applications on heterogeneous clusters

Abstract. Distributed applications running on clusters may be composed of several components with very different performance requirements. The FlowVR middleware allow the developer to deploy such applications and to define communication and synchronization schemes between components without modifying the code. Whereas it eases the creation of mappings, FlowVR does not come with a performance model. Consequently optimiza-tion of mappings is left to the developer skills. This task seems difficult to perform when the number of components and cluster nodes grow. Moreover the cluster may be composed of heterogeneous nodes making this task even more complex. In this paper we propose an approach to predict performances of FlowVR distributed applications given a mapping and a cluster. We also give some advices to the developer to create efficient mappings and avoid configurations which may lead to issues. Since the FlowVR model is very closed to underlying models of lots of distributed codes, our approach can be useful for all designers of such applications.