Skip to Main content Skip to Navigation

Placement d'Applications Parallèles en Fonction de l'Affinité et de la Topologie

François Tessier 1, 2
2 TADAAM - Topology-Aware System-Scale Data Management for High-Performance Computing
LaBRI - Laboratoire Bordelais de Recherche en Informatique, Inria Bordeaux - Sud-Ouest
Abstract : Computer simulation is one of the pillars of Sciences and industry. Climate simulation, cosmology, or heart modeling are all areas in which computing power needs are constantly growing. Thus, how do we scale these applications? Parallelization and massively parallel supercomputers are the only ways to achieve this. Nevertheless, there is a price to pay considering the hardware topologies incessantly complex, both in terms of network and memory hierarchy. The issue of data locality becomes central: how to reduce the distance between a processing entity and data to which it needs to access? Application placement is one of the levers to address this problem. In this thesis, we present the TreeMatch algorithm and its application for static mapping, that is to say at the launchtime of the application, and the dynamic placement. For this second approach, we propose the awareness of data locality within a load balancing algorithm. The different approaches discussed are validated by experiments both on benchmarking codes and on real applications.
Complete list of metadata

Cited literature [91 references]  Display  Hide  Download
Contributor : François Tessier Connect in order to contact the contributor
Submitted on : Thursday, March 29, 2018 - 4:41:12 PM
Last modification on : Monday, November 30, 2020 - 4:18:03 PM


Files produced by the author(s)


  • HAL Id : tel-01753456, version 1



François Tessier. Placement d'Applications Parallèles en Fonction de l'Affinité et de la Topologie. Calcul parallèle, distribué et partagé [cs.DC]. Univ. Bordeaux, 2015. Français. ⟨tel-01753456⟩



Record views


Files downloads