CSCS announced beginning of April the successful deployment of the SLURM Resource Manager on Rosa (Cray XT5).
Slurm was developed by Lawrence Livermore and has been ported to the Cray architecture by the National Supercomputing Services (NSS) team NSS at CSCS. Slurm has been customized with the in-house scheduling algorithm of CSCS. The 1840 node XT5 «Rosa» stays 99.75% full of running jobs.
This deployment is fully redundant with backup accounting and slurm controllers. In addition GPU resource consumption has been implemented on GPU test cluster. External login nodes and environment modules allow to batch to any of our clusters from one central entry point.