Andy Jones has graciously agreed to share his presentation with you guys as well. Andy is a long time member of our community, currently helping out at NAG with HECToR and several other things, and he also does quite a bit of writing in our community as well.
Here is the abstract for his talk
This talk will provide an overview of the UK’s national supercomputing service, HECToR, including the imminent deployment of the world’s first Cray XT6 system. HECToR funds 20 staff to renovate user applications to get significantly more science per dollar. Through examples, the talk will show major increases in application performance at large scales achieved under this programme to date. These successes have substantially reduced the cost of supercomputer time for participating researchers and enable a greater amount of science to be delivered from the government’s existing investment in supercomputers. The talk will also describe how the programme ensures strong engagement with the user community and present feedback from the users community.
Slides from Andrew’s talk, “Who is HECToR and How Did He Help Users Get More HPC for Their Money?” [PDF]
hi andy!
Hi Andy and John,
I’m glad to see HECToR is doing well and it seems NAG has done quite a decent job, but I’d be very curious to know more about the interplay between NAG and the key HPC academic (and to a lesser extent, government) sites in the UK, At my institution, we’re having a hell of a time finding people with decent HPC expertise – a situation that I imagine is equally true in the UK. A colleague at another institution told me their approach is to get students involved early, and if they’re good then offer them a position upon graduation. Having an academic institution like the EPCC or QUB calling the shots would seem to be a boon to student involvement, as well as ensure an unbiased approach towards partnerships with other commercial entities, and they certainly do have the local expertise.
Maybe this is a long game / short game issue. In the short term, professional support from experienced people (NAG experts) offers more immediate benefits to the UK’s HPC strategy, but long term I think having students hands-on with real projects, while maybe taking slightly longer to get results, gets them into the fold and doing real work faster, which translates to more people sticking with it and… well, people like us having an easier time finding talent. Nothing against NAG at all, again, congratulations on the fine work. I just think I’d be slightly concerned if, instead of an NSF grant for GPU-based systems going to, say, ORNL / UT, it was given to nVidia. The latter might have more immediate expertise, but we need open growth and leadership in HPC, especially within the academic community. You don’t form leaders without giving them two things – one, the opportunity to lead, and two, people to lead. 😉
(Note: I’m certainly all for -contracting- with NAG, of course!)
My two strategic cents. Andy, if you’re listening, I’d love to hear your thoughts.
– Brian
The PDF link seems to be broken.
@Patrick – works fine for me; are you still having trouble?
No, it works now. Thanks John.
Looks like the slides are gone again. Link leads to a Not Found error. 🙁
Honestly, with as much trouble as I’m having with this, you’d think I was putting a man on the moon. Typo in the URL, fixed again.
@Brian,
Thanks for your comments. I’d be happy to provide more information on how NAG interacts with key academic sites in the UK (and globally) – please get in touch directly via http://www.nag.com. In the case of the two sites you mention, EPCC and QUB, we do work with them (e.g. EPCC, one of Europes leading HPC centres, is our partner in delivering the HECToR service). NAG, of course, is a not-for-profit spin-out from an inter-university software collaboration 40 years ago, and whilst we survive by selling numerical libraries and HPC consulting and software engineering services, we also remember our heritage and continue to support the academic community – both in research and education terms.
I would like to note that NAG trades on our impartiality – that is we believe ourselves to be at least as unbiased as any academic site when providing HPC advice to our customers and engaging in development projects.
more @Brian,
On the braoder context of your comments, I agree with you about needing to support the future of HPC through investing in our students – indeed InsideHPC covered some of NAG’s efforts in this recently – http://insidehpc.com/2010/03/12/nag-turns-40-and-expands-the-nag-student-awards-program/. However, I am pretty confident that we do NOT want to send a message to students considering including HPC in their studies that the only place they will be able to use HPC skills is academia. They must see a viable set of industry actively involved in HPC – using HPC (e.g. Airbus, Boeing, Proctor & Gamble), providing HPC services (e.g. NAG) and HPC products (e.g. Cray, IBM, Intel, AMD, etc).
The key thing is that I agree with you on fostering growth of our future HPC community – academia has a big role to play, but without visible industry employment options it will be a hard sell as a career for students.