Section: Overall Objectives

Grand-Large General Objectives

Grand-Large was evaluated in september 2012 and was supposed to end in December 2012. Additional information can be found in the 2012 Grand-Large evaluation report. It was continued in 2013 with purpose to create a new team involving some of the Grand-Large members.

Grand-Large is a research project investigating the issues raised by High Performance Computing (HPC) on Large Scale Distributed Systems (LSDS), where users execute HPC applications on a shared infrastructure and where resources are subject to failure, possibly heterogeneous, geographically distributed and administratively independent. More specifically, we consider large scale distributed computing mainly, Desktop Grids, Grids, and large scale parallel computers. Our research focuses on the design, development, proof and experiments of programming environments, middleware and scientific algorithms and libraries for HPC applications. Fundamentally, we address the issues related to HPC on LSDS, gathering several methodological tools that raise themselves scientific issues: theoretical models and exploration tools (simulators, emulators and real size experimental systems).

Our approach ranges from concepts to experiments, the projects aims at:

  1. models and fault-tolerant algorithms, self-stabilizing systems and wireless networks.

  2. studying experimentally, and formally, the fundamental mechanisms of LSDS for high performance computing;

  3. designing, implementing, validating and testing real software, libraries, middleware and platforms;

  4. defining, evaluating and experimenting approaches for programming applications on these platforms.

Compared to other European and French projects, we gather skills in 1) large scale systems formal design and validation of algorithms and protocols for distributed systems and 2) programming, evaluation, analysis and definition of programming languages and environments for parallel architectures and distributed systems.

This project pursues short and long term researches aiming at having scientific and industrial impacts. Research topics include:

  1. the design of middleware for LSDS (XtremWeb and PVC)

  2. large scale data movements on LSDS (BitDew)

  3. fault tolerant MPI for LSDS, fault tolerant protocol verification (MPICH-V)

  4. algorithms, programming and evaluation of scientific applications LSDS;

  5. tools and languages for large scale computing on LSDS (OpenWP, YML).

  6. Exploration systems and platforms for LSDS (Grid'5000, XtremLab, DSL-Lab, SimBOINC, FAIL, V-DS)

These researches should have some applications in the domain of Desktop Grids, Grids and large scale parallel computers.

As a longer term objective, we put special efforts on the design, implementation and use of Exploration Tools for improving the methodology associated with the research in LSDS. For example we had the responsibility of the Grid eXplorer project founded by the French ministry of research and we were deeply involved in the Grid5000 project (as project Director) and in the ALADDIN initiative (project scientific director).