Grid'5000 user report for Arnaud Lefray
User informationArnaud Lefray (users, user, lyon, ml-users user)
More user information in the user management interface.
- Intended Grid'5000 usage for user Arnaud Lefray (Ecole Polytechnique Fédérale de Lausanne - Laboratoire des Systèmes Répartis) (Middleware) [achieved]
Description: Working on Fault-Tolerance on Petascale/Exascale machines, the goal is to evaluate experimentally different implementations of fault-tolerant protocols on large scale HPC applications more particularly based on replication. Applications benchmark are written in C using MPI. The experiment lasts ~15 min for each application (~5).
- Replication in OpenMPI (Middleware) [planned]
Description: The goal is to evaluate an implementation of replication in OpenMPI at large scale. The evaluated applications are the NAS Parallel Benchmarks and few real HPC applications (not decided yet). We mesure : - The replication protocol overhead. - The impact of buffering instead of synchronization