Team GRAAL

Members
Overall Objectives
Scientific Foundations
Application Domains
Software
New Results
Other Grants and Activities
Dissemination
Bibliography

Section: Other Grants and Activities

National Contracts and Projects

INRIA Grant: Software development for MUMPS (“Opération de Développement Logiciel”)

Inria has been financing Aurélia Fèvre, on contract from September 1, 2005 to August 31, 2007, as an engineer to work on the development of the Mumps software package. This year, Aurélia has mainly worked on extending the test suite of Mumps (non regression tests and improvement of the code coverage).

French ministry of research grant: Grid'5000, 3 years, 2004-2007

ENS Lyon is involved in the GRID'5000 project [57] , which aims at building an experimental Grid platform gathering nine sites geographically distributed in France (17 laboratories). Each site hosts several clusters connected through the RENATER network.

Graal is participating in the design of the École normale supérieure de Lyon node. The scalability of Diet together with several scheduling heuristics will be evaluated on this platform.

ANR grant: ALPAGE (ALgorithmique des Plates-formes À Grande Échelle), 3 years, 2005-2008

The goal of this project is to gather researchers from the distributed systems and parallel algorithms communities in order to develop efficient and robust algorithms for some elementary applications, such as broadcast and multicast, distribution of tasks that may or may not share files, resource discovery. These fundamental applications correspond to the spectrum of the applications that can be considered on large scale, distributed platforms.

Yves Robert is leading the Rhône-Alpes site of this project, which comprises two other sites: Paris (LIX and LRI laboratories) and Bordeaux-Rennes (Paris and Scalapplix projects). Anne Benoit and Frédéric Vivien participate in this project, together with Lionel Eyraud, who held a post-doctoral position until August 31.

ANR grant: Stochagrid (Scheduling algorithms and stochastic performance models for workflow applications on dynamic Grid platforms), 3 years, ANR-06-BLAN60192-01, 2007-2010

Grid computing platforms and components are subject to a great variability. Statistical models are mandatory to deal with changes in resource performance, such as CPU speeds or link bandwidths. Traditionally, Markov chains are used to capture the inherent uncertainty linked to parameter estimation. However, Markov chains lack a key feature: because they are memoryless, they cannot accurately model the performance of parallel systems periodically interacting through message exchanges in steady-state mode.

In contrast, sophisticated static scheduling strategies have been developed to map workflow applications on static Grid computing platforms. Optimal algorithms have been designed to map simple pipeline skeleton kernels onto heterogeneous clusters and Grids. Such applications operate in pipeline mode, and standard objective functions include maximizing the throughput and/or minimizing the response time (latency), for each data set.

A major goal of this project is to fill the gap between both approaches. On the one hand, statistical models are mandatory to account for the variability and dynamicity of resources. On the other hand, efficient scheduling algorithms only exist for static, dedicated platforms. We need a new stochastic model able to capture the performance of dynamic parallel systems accurately. This new model will be non-Markov for system interaction but will be Markov-based for platform characteristics (fault-tolerance and variability). The design and evaluation of this new model will be the first key contribution of the project. New, robust, scheduling algorithms will be designed and evaluated on top of this model, thereby providing the first stochastic testbed for workflow applications on Grid platforms. The third key contribution of the project will be the design of a prototype library for deploying workflow applications on computational Grids.

The project is entirely conducted within the GRAAL team (Anne Benoit and Yves Robert are the permanent members involved).

ANR grant CICG-05-11: LEGO (League for Efficient Grid Operation), 3 years, 2006-2008

The aim of this project is to provide algorithmic and software solutions for large scale architectures; our focus is on performance issues. The software component provides a flexible programming model where resource management issues and performance optimizations are handled by the implementation. On the other hand, current component technology does not provide adequate data management facilities, needed for large data in widely distributed platforms, and does not deal efficiently with dynamic behaviors. We choose three applications: ocean-atmosphere numerical simulation, cosmological simulation, and sparse matrix solver. We propose to study the following topics: Parallel software component programming; Data sharing model; Network-based data migration solution; Co-scheduling of CPU, data movement and I/O bandwidth; High-perf. network support. The Grid'5000 platform provides the ideal environment for testing and validation of our approaches.

E. Caron is leading the project, which comprises six teams: Graal /LIP (Lyon), PARIS/IRISA (Rennes), RUNTIME/LaBRI (Bordeaux), ENSEEIHT/IRIT (Toulouse), Cerfacs (Toulouse) and CRAL/ENS-Lyon (Lyon). A. Amar, R. Bolze, Y. Caniou, F. Desprez, J.-S. Gay and C. Tedeschi also participate in this project.

ANR grant ANR-06-CIS-010: SOLSTICE (SOlveurs et simulaTIon en Calcul Extrême), 3 years, 2007-2009

The objective of this project is to design and develop high-performance parallel linear solvers that will be efficient to solve complex multi-physics and multi-scale problems of very large size (10 to 100 millions of equations). To demonstrate the impact of our research, the work produced in the project will be integrated in real simulation codes to perform simulations that could not be considered with today's technologies. This project also comprises LaBRI (coordinator), Cerfacs , INPT-IRIT, CEA-CESTA, EADS-CCR, EDF R&D, and CNRM. We are more particularly involved in tasks related to out-of-core factorization and solution, parallelization of the analysis phase of sparse direct solvers, rank detection, hybrid direct-iterative methods and expertise site for sparse linear algebra.

Emmanuel Agullo, Aurélia Fèvre and Jean-Yves L'Excellent participate to this project.

ANR grant ANR-06-MDCA-009: Gwendia (Grid Workflow Efficient Enactment for Data Intensive Applications), 3 years, 2007-2009

The objective of the Gwendia (http://gwendia.polytech.unice.fr/doku.php )project is to design and develop workflow management systems for applications involving large amounts of data. It is a multidisciplinary project involving researchers in computer science (including GRAAL) and in life science (medical imaging and drug discovery). Our work consists in designing algorithms for the management of several workflows in distributed and heterogeneous platforms and to validate them within DIET on the Grid'5000 platform.

SEISCOPE Consortium

The SEISCOPE project coordinated by Geosciences Azur focuses on wave propagation problems and seismic imaging. Our parallel solver Mumps has been used extensively for finite-difference modeling of acoustic wave propagation (see publication  [Oops!] ). We also started using the large-scale test problems arising from this project to design and experiment our out-of-core approaches. The SEISCOPE project is supported by ANR (Agence National de la Recherche Française), and by BP, CGG, SHELL and TOTAL.

Emmanuel Agullo and Jean-Yves L'Excellent participate to this collaboration.


previous
next

Logo Inria