GRAAL

see : http://graal.ens-lyon.fr/index_us.html

Parallel computing has spread into all fields of applications, from classical simulation of mechanical systems or weather forecast to databases, video-on-demand servers or search tools like Google. From the architectural point of view, parallel machines have evolved from large homogeneous machines to clusters of PCs (composed of nodes, each housing multiple processors and a shared memory, these nodes being connected by high speed networks like Myrinet). However the need of computing or storage resources has continued to grow, leading to the need of resource aggregation through Local Area Networks (LAN) or even Wide Area Networks (WAN). The recent progress of network technology has made it possible to use highly distributed platforms as a single parallel resource. This has been called Metacomputing or more recently Grid Computing. While an enormous effort has recently been devoted to study this important subject, we believe that many projects have failed to study fundamental problems such as the complexity of problems and algorithms, and scheduling heuristics. Also theoretical results are rarely validated on available (software) platforms.

There are two main challenges for the widespread use of heterogeneous distributed platforms: the development of environments that will ease their use (in a seamless way) and the design and evaluation of new algorithmic approaches for applications using such platforms. Aims of the GRAAL project

In the GRAAL project we work on the following research topics:

• algorithms and scheduling strategies for heterogeneous platforms and the Grid,
• environments and tools for the deployment of applications in a client-server mode.

The main keywords of the GRAAL project are: Algorithmic Design, Middleware/Libraries, Applications over heterogeneous architectures and the Grid. GRAAL is a project of LIP (Laboratoire de l'Informatique du Parallélisme) jointly operated by CNRS, INRIA and École normale supérieure de Lyon. This project is located in Lyon. Research themes

• Scheduling strategies and algorithm design
• Parallel sparse direct solvers
• Network enabled servers

International and industrial relations

• International collaborations in academia (University of Tennessee, Knoxville (UT), University of California San Diego (UCSD), Technical University of Munich, and University of Bergen).

Software

DIET, FAST, MUMPS, SimGrid 2

Scientific leader : Frédéric VIVIEN, Frederic.Vivien@inria.fr +33 4 72 72 84 96

Secretary : +33 4 72 72 80 37