I was cleaning up some files and I found this time lapse that I did when we were building Kelvin, it’s a few years old now. Even by current standards it’s still pretty respectable.
We had to unpack and install all the infiniband cards ourselves, cabled, racked, installed, configured and burnt it in for production usage. The cluster has 100 nodes, each node has 2 sockets, each socket has 6 cores and 24gb of ram.
The most interesting thing about this machine was that we got it with QLogic infiniscale/infinipath HCA’s and switching. It was one of the if not the lowest latency networking that we could get at the time.
When we were configuring this system we had also redesigned the storage system roughly around this period and went with a cluster to cluster GPFS configuration.
Oh and we threw out the old cluster which was about 10racks of IBM e326 machines in the process.
Share!
See Also
- supercomputing 2012 and the arms race to exascale
- digital preservation and archiving is a hpc problem
- slurm bank that big script for banking in slurm
- cports for building applications and libraries for hpc systems
- crowbar for deploying systems