Peer to peer large scale linear algebra programming and experimentations

  • Authors:
  • Serge G. Petiton;Lamine M. Aouad;Laurent Choy

  • Affiliations:
  • Laboratoire d'Informatique Fondamentale de Lille, and INRIA Futurs, University of Science and Technology of Lille, France;Laboratoire d'Informatique Fondamentale de Lille, and INRIA Futurs, University of Science and Technology of Lille, France;Laboratoire d'Informatique Fondamentale de Lille, and INRIA Futurs, University of Science and Technology of Lille, France

  • Venue:
  • LSSC'05 Proceedings of the 5th international conference on Large-Scale Scientific Computing
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper we discuss the deployment of large scale numerical applications on a Grid. Currently, Grid deployments have coarse-grained and massively parallel nature and are suited for applications that are not communication intensive. For our numerical applications, we minimize the communication needs by a good choice of data placement, in particular by using persistent storage of data. We also introduce out-of-core programming for the task farming paradigm in order to achieve interesting degree of scalability. We discuss the performances of the bisection method to compute the eigenpair of a real symmetric tridiagonal matrix and a block-based matrix-vector product. As experimental middleware we use the XtremWeb system in different configurations; a local one and a WAN based platform with heterogeneous resources distributed on two geographic sites: the university of Lille I and Paris-XI university in Orsay.