At Virtual SC20: How to Build a Supercomputer – Supermicro Showcases 3 LLNL HPC Clusters

At Virtual SC20, we sat down with Mike Scriber, he’s Supermicro’s senior director, server solution management, to talk about three supercomputing clusters the company has installed at Lawrence Livermore National Laboratory. They are the 1500-node ‘Ruby‘ (targeting COVID-19 research), the GPU-dense Corona (the name pre-dates the current pandemic) and the memory-rich ‘Mammoth‘.

Supermicro, which specializes in servers, storage, blades, rack solutions, networking devices, server management software and high-end workstations, has extensive experience in integrating HPC systems, experience that was called upon for the three clusters Scriber discusses in this interview.

Of the demanding cooling requirements of the Ruby cluster, Scriber said this of Supermicro’s TwinPro liquid cooling technology:

We got about 64 nodes in each rack, so that gave us great density. With that, usually, we have some thermal issues, not so much within the system but thermals for the air conditioning system. But with liquid cooling, we were able to cool the vast majority of their thermals through liquid cooling, which gives you really, really great efficiencies, and also allows you to have great density.

Visit this site for more information about Supermicro at SC20.