Clusters Home

Context

The Sophia Antipolis research unit of the French National Institute for Computer Science and Automation (INRIA) has an experimental and evolutive computing platform.

This platform relies on a cluster computing approach : several generations of high performance servers are combined to provide heterogeneous parallel architecture. These computational nodes are arranged in sub-clusters each of them being associated to a particular interconnection technology.

This platform is evolutive in the sense that computational nodes will change in numbers and types in order to follow the rapid technology evolution for servers and high performance networks.

Usage

The platform is open for all teams of the INRIA Sophia research center. It can be used for both parallel computations, GPU computation and sequential jobs.

History

2014-07-17

16 new machines added nef013-028 (320 new cores)

2013-11-21

new NFS server for home directories (7TB available instead of 1.3TB)

2013-06-18

6 new machines added nef007-012 (384 new cores), PGI 13.5 installed.

2013-04-22

memory upgrade on nef001-006 (from 128GB to 256GB)

2012-12-06

software upgrade (fedora 16, PGI 12.6, CUDA 5.0, ...)

2011-07-07

software upgrade (fedora 14, PGI 11.5, CUDA 4.0, ...)

2011-02-03

8 new machines added, major software upgrade (Fedora 12, PGI 11.1, openmpi 1.4.3 ...). QDR Infiniband network instead of myrinet

2010-03-12

software upgrade (openmpi 1.4.1, PGI 10.3, ...)

2009-06-09

major software upgrade (from Fedora 7 to Fedora 10)

2009-02-20

GPU cluster available for tests : two nodes with Fedora 10 and 4 Nvidia GPUs

2008-02-28

19 new machines added, minor software upgrade (PGI 7.1 and openmpi 1.2.5)

2007-08-16

major software upgrade (from Rocks to Fedora 7)

2007-01-30

32 new nef machines added

2005-06-20

Slides on How to use Sophia's cluster are available

2005-03-09

The nef cluster is online! This platform includes one frontend and 32 compute nodes (dual opterons)