Grid'5000 user report for
User information( user)
More user information in the user management interface.
- MPI'5000 (Middleware) [in progress]
MPI5000 is a new transparent layer placed between MPI and TCP allowing application composed of several tasks to be correctly distributed on available node regarding the grid topology and the application scheme. Thus, our layer needs two data files: a file describing the grid topology including available nodes, both latency and bandwidth between the nodes and between sites; another file describing the application communication patterns with the size and the amount of messages sent between MPI processes. Using these two informations, our layer should realise an efficient placement of tasks on grid nodes.
Our layer also propose to transparently slipt TCP connections between MPI processes in order to take into account the grid topology. This new architecture is based on a system of relays placed at the LAN/WAN interface. We replace each end-to-end TCP connection by three connections (two on the LAN between a node and a relay, one on the WAN between two relays). Thus, we expect a faster lost recovery on LAN as well as a reduction of memory used because for local TCP buffers (they depend on RTT latency of a connection). On the relay, we planned to use different TCP implementations or different protocols for local and distant communications. The relays could also implement a different scheduling strategy of the messages in function of the data size, for example we could give priority to small message (usually control messages). Finally, as MPI applications are mostly using small messages, they are more penalised if the network is congestionned by large flows. We planned to reserve bandwidth in order to optimise MPI communications on the long distance shared link. The implementation of our proposition is based both on a library between MPI and system calls and relays daemon. Thus, the architecture is independant of MPI implementations.
Results: For the moment, relays and library are in a test phase. We are now testing our architecture in Grid'5000. Finally, we will implement the optimisations proposed previously.
- Optimization of Long-distance communications for MPICH-Madeleine (Networking) [achieved]
Description: To be executed in a grid, applications needs a support like MPI. But MPI was created for clusters. In grids, there is at least two more constraints to manage: heterogeneity and long distance management.
MPICH-Madeleine manage heterogeneity properly but no experiments was done with long-distance.
I use Grid'5000 to see how MPICH-Madeleine behave in the grid. I compare local perfomances and long distant ones in order to adapt this implementation for the grid. Principally, I try to optimize long-distance communications.
Results: Optimizations done lead to obtain a bandwith of 600 Mbps instead of 95 Mbps for sending MPI messages over a Wide Area Network.
- MPI'5000 (Middleware) [planned]
Description: MPI'5000 aims to be an implementation of the Message Passing Interface standard within a grid environment such as GRID'5000. Our goal is to optimize communication software stack from the application level down to the hardware level. For instance, theses optimizations can take place as well as in MPI as in high-speed networking stacks or standard networking stacks such as MX/Myrinet, VERBS/Infiniband, TPORTS/ElanLib/QSNet, TCP/IP/GigaEthernet, … Indeed, the use of computing grids such as GRID’5000 requires high speed communications in an heterogeneous world : GRID’5000 is composed of several heterogeneous clusters interconnected by a fast, wide and dedicated area network ; these clusters are themselves interconnected by a specific high speed network such as Myrinet, Infiniband, QSNet, 10 Gigabit Ethernet… With MPI’5000, we want to test new mechanisms in the context of computing grids improving communications between all these heterogeneous networking technologies.
- R\'eseau longue distance et application distribu\'ee dans les grilles de calcul : \'etude et propositions pour une interaction efficace  (national)
EntryType: phdthesis Author: Ludovic Hablot School: ENS Lyon, Universit\'e de Lyon
- Etude d'impl\'ementations MPI dans une grille de calcul  (national)
EntryType: inproceedings Author: Hablot, Ludovic and Gl\"uck, Olivier and Mignot, Jean-Christophe and Vicat-Blanc Primet, Pascale Booktitle: Actes de Renpar'08 Month: F\'evrier
- Comparison and tuning of MPI implementation in a grid context  (national)
EntryType: inproceedings Author: Hablot, Ludovic and Glück, Olivier and Mignot, Jean-Christophe and Genaud, Stéphane and Vicat-Blanc Primet, Pascale Booktitle: In Proceedings of 2007 IEEE International Conference on Cluster Computing (CLUSTER) Month: September Pages: 458-463