All the CIMENT platforms are integrated into a local grid that aims at optimizing the resources usage. Actualy, the load of the computing clusters of such a local computing center is very irregular: one can be overloaded while at the same time one is completely free, because there are no experiments at this time.
Since 2000, CIMENT allows every users of parametric applications to launch their jobs on idle processors of every computing clusters of the grid, even for users that are not "legitimous" on a particular project supercomputer. This is made possible by the use of a specific middleware CIGRI and resource manager OAR that deal with a special job type called "best-effort jobs".
CiGri is the middleware used by the CIMENT grid. It offers a simple interface to users that have to launch big set of parametric tasks. It is very efficient with best-effort jobs and manages the re-submissions of such jobs.
CiGri is mainly developped by CIMENT and the LIG. The current version (3.0) is developed with the collaboration of Grid5000.
You'll find more informations about CiGri on it's page.
Storage grid using IRODS
The CIMENT storage infrastructure consists of several storage nodes installed on different sites as near as possible of each CIMENT supercomputer. There's currently a total storage capacity of 700TB, on several nodes located on 3 different sites (SIMSU/Campus, INRIA/Montbonnot and CECIC/Campus). Some of the clusters have a 10Gb/s connection to the local storage nodes. It means that some computing nodes can upload/download files up to 800MB/s!!. The storage is operated by IRODS (https://www.irods.org). It is a grid storage.
Grenoble also hosts a Grid5000 site. The computers of this site are accessible by CIMENT users through the CIMENT grid. This site is also used for some trainings, especially for GPU-computing.
More generally, there are some frequent interactions between CIMENT and GRID5000, with system administration skills exchanges.