Our very own Mini-Supercomputer
We strongly believe that in these modern times computational and experimental studies have to go hand in hand. You could say, that’s the raison d’etre for this lab. No computational simulation is worth anything, if it doesn’t agree with reality and modern experimental studies are more often about validation of computational models than not.
With that in mind, we have been working on our computational capability to compliment the experimental side, and have set-up ecmBEO, a small beowulf/warewolf cluster.
This cluster is built from decommissioned HP Z-Stations operating a CentOS 7 Linux, warewolf cluster distribution software with stateless and headless slave nodes (in the computational context, the term “slave” is still politically and, more importantly, technically correct), slurm load leveller and a variety of computational software (to be extended).
To stick with the theme, we have put a couple of Ironwolf HDDs in, which provide a workspace of 8TB (mirrored).
This cluster is not intended as competition to the SHU cluster, but as a prototyping system, with lower barrier to entry – especially for students. Since we are using the same load leveller system (slurm), all scripts and run-files should be easily transferable to the big machine.
So, if you are able to operate a numerical code from the command line, and want to give our pack of wolves a workout, get in touch with us.
If you have a Z210 or newer workstation that is not used anymore, please donate it. If you are due a system upgrade, ask IT if you can add the old machine to the cluster.