Hierarchical adaptive multi-mesh partitioning algorithm on heterogeneous systems

Abstract

Heterogeneous systems are increasingly being used as platforms for resource-intensive distributed parallel mesh applications. A critical contributors to the performance of such applications is i) the scheduling of constituent application tasks ont he system and ii) the mesh partitioning. Since often application submissions from users cannot improve the total turnaround time, the application must be scheduled automatically by the scheduler using information about its scalability characteristics. To obtain the best performance, it is imperative to take into account both application-specific and dynamic system information in developing i) a schedule which meets his performance criteria and ii) a mesh partitioning scheme that takes into account this system heterogeneity. In [1], we have presented a mesh partitioning scheme, that takes into account the heterogeneity of CPU and networks. Load balancing mesh partition strategy improves the performance of parallel applications running in a heterogeneous environment. In this work, we present a new hierarchical adaptive scheme that allows to optimize the scheduling of N parallel mesh applications simultaneously on a heterogeneous system. The new scheme consists in selecting the partition size of each application and then partitioning each application on the allocated heterogeneous ressources. We illustrate our scheduling approach with a detailed description and results for a distributed 3D CFD application on a heterogeneous platform. © 2010 Springer.