Team presentation

Project-team GRAND LARGE works in the framework of large scale computing and the Globalization of Computer Resources and Data (GRID) initiative. The project is especially interested in large scale parallel and distributed systems supposed to work over the Internet, intranets, LANs or broadband networks. The project’s approach concerns middleware and low-level programming environments, between low-level system mechanisms and high-level programming environments.

The team first strives to study experimentally as well as formally the basic mechanisms of such systems, in the limited framework of computing, that is to say the hard scientific points and technological deadlocks. The team also designs, develops, validates and tests software modules and actual platforms. Finally, GRAND LARGE works on defining, assessing and experimenting on the programming of such platforms.

GRAND LARGE differs from other European and French projects by an original approach based on competence in scaling (large scale diffusion, resistance to volatility, heterogeneity, inter administration domain security, and so on) acquired during the XtremWeb projects (LRI, cluster and grid), in the design and formal validation of algorithms and protocols for distributed systems (LRI, parallel computing), in programming, assessing, analyzing and defining languages and programming environments for parallel architectures and distributed systems (LIFL, parallel methods and algorithms).

The team carries out short and long term studies, aiming at scientific and industrial results. Its research themes concern security, volatility management, large scale scheduling, data circulation, and parallel and distributed programming. More generally, the research developed by the project have applications in very large parallel systems (equipped with several thousands processors) and grids (PC grids, cluster grids). In the long run, the conditions for a merger of the Grid and Peer to Peer systems will be studied, by integrating experimentation in a controlled environment, in addition to traditional basic and experimental techniques.

Research themes

A) Construction/experimentation of a global, peer to peer computing environment over the Internet, Intranet and LAN.

• Organization of the connection system
• Security
• Data storage and circulation
• Simulation and comparison of centralized, hierarchical and entirely distributed scheduling mechanisms
• Extension of MPICH-V (fault tolerant infrastructure for complex communication protocols)
• The problem of GRID-P2P merger

B) Dealing with volatility and reliability at a basic level: adequation of self-stabilization, network level protocols, verification

• Reliability
• Study of network level protocols
• Protocol verification

C) Definition of programming and numerical methods for large scale computing

• Algorithms, programming and evaluation of large scale peer to peer grids
• Tools and languages for large scale computing

International and industrial relations

National collaborations:

• ACI GRID "GRID2", "DataGraal", "CGP2P",
• Projet Pluri Formations (Université Paris XI), "Calcul Global : Augernome XtremWeb",
• CNRS "Dynamo" Specific Initiative,
• CNRS JemSTIC "Mobicoop" Initiative,
• CNRS JemSTIC "STAR" Initiative,
• Minsitry of Research "XtremWeb" Specific Initiative,
• RNTL "cASPer" project,
• CCEA DAM contract

International collaborations:

• Bilateral PICASSO collaboration "Algorithmics of mobile agent systems"

• GRAND-LARGE participates to the QosCosGrid european project

Industrial collaborations:

• GIE EADS,
• IFP,
• CEA