Notes

Our Notes on Running HPC Services

Here we compile our notes from experience delivering managed services and looking after a variety of HPC & AI environments.

We write these notes when we have something worth saying and update them as our thinking moves on. How we run an HPC/AI service is not a secret recipe, it’s an approach which we can deliver to you or help you in using yourself.

The bottom line is that:

  • Control is a choice. Managed doesn’t mean that you have to hand over the keys to your service.
  • Health is not just hardware. Scheduler throughput, user experience and service evolution are all parts of your service health.
  • Process matters. A repeatable approach to identifying, quantifying and addressing your service’s health status is just as important as the action taken

If a note lands close to a problem you have, view our offerings or email us and we’ll work out where you sit on the dial.