HPC is a Broad Niche

A few weeks ago, I asked what good resources are available for learning about HPC. I had been asked the question myself several times in the prior weeks as we had been interviewing for a junior HPC role.

At the time, I didn’t have a great answer. Even now, with all your help, while it is better and has at least a couple of resources I can point to, it’s not amazing (HPC Catalog has been updated to include these). One of the things I feel is really missing is a primer. A short intro to HPC that just covers the basics. More on this at the end

I think part of the problem is that whilst HPC is a fairly niche topic, it is one that is incredibly broad. You could spend your entire career optimising  BLAS or math libraries for use in HPC but have little exposure to high performance networking or schedulers. Yet all those things lie within the domain of HPC.

Part of the challenge I think, is simply, which bit of it do you want to learn about. Without even attempting to be an exhaustive list and just off the top my head the following topics would all fall under the HPC umbrella

• Job and resource scheduling
• MPI
• CPU architecture
• Networking
• Storage and parallel file systems
• Linear algebra and the BLAS libraries

I guess if you already know which bit of HPC you want to learn about it would probably be easier to find more resources about it. But how do you even know that those are all things within the world of HPC? I guess the roadmaps.sh approach proposed by Alex might be good for that?

The picture? Yea its mostly a joke… but I have a non negotiable need to train up a junior HPC engineer soon, so I started putting something together. I figured it would just be another blog/LinkedIn post. I’m at 10 chapters so far… and not quite sure what I do with it!