HPC Slurm Cluster

HPC Slurm

What it does.

The HPC Slurm cluster is our general workhorse HPC system. It provides dedicated user, group and scratch data areas and the software installations are managed in user-land. This is an heterogeneous resource server farm, with a mix of AMD Opteron 6134, 6174, 6272, 6278 and Intel E5-2603, E5-2660 CPUs. The nodes have a mixed configuration of 128GB and 256GB of RAM with 94 and 14 nodes respectively.

How we use it.

This heterogeneous distributed memory resource is used on a daily basis by EI’s research groups and P&P to process the large volumes of data generated from the sequencing machines.

The cluster manages about 39.417 jobs weekly, the typical job requires a large amount of memory and less than 50 CPUs, our top users require up to 1.387.345 CPU hours in a week. All these jobs are accessing and processing the 2+ PB of data available on the EMC Isilon storage solution.

Our SLURM resource manager is configured using a fairshare algorithm which ensures that the computing resources are used effectively while at the same time keeping the usage fairly distributed over all the users.

Related projects.

Platforms & Pipelines, Research groups, and every HPC user at EI.

Related reading.