High Performance Cluster now online using SLURM

Wed, 10/05/2016 - 11:02am -- root

All,

Statistics is now running our own high performance compute cluster consisting of six - many core, high memory nodes. The cluster will eventually replace the bigmemXX servers and the open, un-restricted way of running R, MATLAB, and other jobs. BigmemXX servers will be considered deprecated and will no longer be monitored for efficiency (if there was any).

The new HPC cluster is considered in BETA as we continue to test and adjust the configuration. The system uses a job scheduler called SLURM (http://slurm.schedmd.com/) which is much like Condor in that it uses simple submit scripts to start and run jobs. The scheduler will allocate the requested resources and queue your jobs so that it will run well with other users.

More information regarding the new Stats HPC (BETA) Cluster and SLURM can be found on the department website here: http://www.stat.wisc.edu/services/hpc-cluster

Please consider giving the HPC cluster a try as we move forward with its full implementation and the retiring of the bigmemXX servers. Knowledge of SLURM submit scripts will be mandatory but the lab will do its best to assist along the way and increase documentation found under the Services tab on our website.

Thanks,
Mike
Statlab

Posted by: 
Mike Cammilleri