The Hpcomputosaurus-rex

Let’s be honest. HPC is a bit of a dinosaur. An Hpcomputosaurus-rex. The h and p are silent by the way. 

Oh, HPC still tries to hang out with the cool younger kids. HPC and AI are often lumped together but let’s be real, it’s a bit like grandad taking his teenage granddaughter out for ice cream. And if you’ve read my musings before you’ll know I say this as afficionado of HPC. Arguably a dinosaur myself, but a fan of HPC non the less.

OpenAI has generated more drama, posts on here and hearts on X just this week than all HPC topics probably have in the last year combined.

I’m not just having a dig at HPC for the sake of it though. The absence of “shiny” in the HPC space has meant that HPC applications have been underinvested in for an extended period of time. We’re still using essentially the same technology to run our HPC clusters (especially in #FinancialServices) that we were 20 years ago. 

The same schedulers that we’ve used for many years, lashed together with bits of Python or Powershell so that they can work on the #cloud. The same schedulers with a variety of bolted on data caching. The same schedulers running on heterogenous hardware they have no comprehension of.

Isn’t it about time we changed that? Else HPC ends up like the poor dinosaur below. Tell me I’m wrong.