Brussels / 2 & 3 February 2019

schedule

Interview with John Garbutt
Square Kilometre Array and its Software Defined Supercomputer. ... and a very fast parallel file system

Photo of John Garbutt

John Garbutt will give a talk about Square Kilometre Array and its Software Defined Supercomputer. ... and a very fast parallel file system at FOSDEM 2019.

Q: Could you briefly introduce yourself?

I am John Garbutt. I have been working on OpenStack related things since December 2010, originally with Citrix, then Rackspace, and now StackHPC. With StackHPC I am looking more generally at the HPC area. I spend much of my non-work time with my one-year-old son. I also play the tuba, hence my IRC nick of johnthetubaguy.

Q: What will your talk be about, exactly? Why this topic?

The Square Kilometer Array (SKA) is actually two radio telescopes, currently in the final stages of prototype work. The SKA’s Science Data Processors (SDP) is facing data processing and flexibility challenges that really push the current capabilities of HPC hardware and software. I have recently been working as a subcontractor for Cambridge University on both the Performance Prototype for the SDP and helping write the recently reviewed architecture, and to share our experience.

Q: What do you hope to accomplish by giving this talk? What do you expect?

Firstly, I hope it’s going to be an interesting look at what is possible using an interesting collection of existing FOSS tools. Secondly, the hope is find others with similar challenges to help get engaged in the OpenStack Scientific Special Interest Group and help work on these problems together.

Q: What are the biggest challenges in building fast storage with OpenStack and related FOSS tools?

The biggest challenge has been integration. Taking all the various pieces and making them work together, providing a stable platform so the scientists can stay productive.

Q: Which capabilities or features are lacking in OpenStack for HPC workloads?

OpenStack is already ready for the majority of HPC workloads. The current work is really about making its operations more efficient (such as work on Ironic Deploy Templates) and looking to make more efficient use of the hardware (such as work on Preemptible instances, Hierarchical Quotas and Blazar Reservations).

Q: Have you enjoyed previous FOSDEM editions?

I have only been lucky enough to previously attend FOSDEM 2016, when I presented in the IaaS and Virtualization dev room. I am certainly finding weekend travel harder after becoming a dad a month or so before last year’s FOSDEM, but at least I can get a train all the way here! I don’t ever remember attending an event with such an epic amount of choice of amazing sessions to attend, never mind in a city with such beauty and beer!

Creative Commons License
Creative Commons License

This interview is licensed under a Creative Commons Attribution 2.0 Belgium License.