Event Details


Lessons learnt federating OpenStack powered Supercomputers

We have recently been working with STFC's IRIS (https://www.iris.ac.uk) federation, adding Cambridge University's Cumulus Supercomputer (87th in November 2018 Top 500 https://www.top500.org/system/179577) into the IRIS federation, and helping other IRIS sites wanting to do something similar. Firstly, we have starting using OpenStack to manage Cumulus.

Exposing these resources using OpenStack have helped provide access to resources in a more flexible way, which has allowed more diverse communities to share the same pool of hardware resources, avoiding the need for costly and unreliable custom dedicated clusters.

Some communities have found Magnum created Kubernetes to be a useful building block. We have used Keystone's Federation, along with Keycloak, to reduce the time spend on user management.

Looking to the future, I talk about what will be needed to help better meet the needs of the STFC IRIS federation.


What can I expect to learn?

Many organisations are looking at how they can more effciently provide compute and storage resources to their users. An OpenStack powered supercomputer can really help server a wide variety of applications from a single pool of resources. This helps users get what they need, and also helps the resource provider more easily manage these more diverse workloads.

Come to find out more about the bumps on the road, and how to avoid them.

Monday, April 29, 2:00pm-2:40pm
Difficulty Level: Intermediate
StackHPC, Principal Engineer
John is a Principal Engineer working on the convergence of OpenStack and HPC with StackHPC. John has been involved with OpenStack since December 2010, and is currently a Nova core reviewer. Between April 2016 and April 2018 he served as a member of the OpenStack Technical committee. He was Nova PTL for the Liberty and Mitaka releases. After studying Computer Science at the University of... FULL PROFILE