Well, this is not yet the info about my 24 core cluster I promised earlier. However, independently, someone else came up with a similar idea of housing an HPC cluster...
Check out Jason's blog, which I found just yesterday... He uses the same cheap IKEA storage my cluster is housed in. However, to me, it does not look like Scrappy Cluster 2.0 is anyway near production, according to the photographs.
Jason writes about starting all nodes at the same time. Not sure about that. In my system, the head node needs to be booted first, the workers will than be sequentially booted. Since the OS and everything else comes from the same host, the head node, through the same network, it actually seems not be a good idea to boot all workers simultaneously.
As a hint and appetizer, I use the "power back" function of the BIOS to switch on the individual nodes.
I hope Jason will pick up the project at some stage again.
