Supercoming Asia 2024 and More
Feb. 23rd, 2024 06:12 pmThe last week was spent in Sydney for Supercomputing Asia 2024, overlooking Darling Harbour with some impressive storms that made me feel good about being inside. With over 1,000 attendees it is obviously smaller than Supercomputing (US) or the International Supercomputing Conference (Germany, my preferred choice), but there was a very good sense of intimacy at this size. Conferences of this sort do have their share of high-level presentations along with some managerial-level buzzwords of where they think the industry is heading, which are sometimes less true than they would like the dollars to follow. "Artificial intelligence" and "quantum computing" were two particularly notable phrases this year, although with regard to the latter, I do note that more careful minds point out the difference between quantum computing (typically conducted on stock-standard HPC systems) and quantum computers (which are still, after many years, a bit of an experimental oddity at best).
The Conference was especially good for getting a sense of the power of some of the truly big iron. Representatives of the two largest systems, Frontier (no 1), Aurora (no 2), and Fugaku (no 4), were present. It is humbling to see systems that have more nodes than cores in the UniMelb system (which is no laggard), and pleasing to get a good grasp of the architecture. I was quietly pleased that my own comments about HPC microprocessor trends from 18 months ago were quite accurate, and I have a special nod to the Thai supercomputing centre which has really become an impressive national facility in its own right. A nice quirk of the conference was that the voice of Siri, Karen Jacobsen was the MC for the plenary sessions. There was, of course, the opportunity to talk to many vendors who are all doing some very interesting things and making plentiful offers, although I do note that there is often a pitch in favour of monopolisation ("we'll do everything for you") rather than interoperability.
For my own part, I presented a talk the "HPC Certification Forum: An Update", which was very well received (particularly the idea of using the skill tree as a teaching structure), and the poster "HPC Training Generates HPC Results". The evening hosted by Xenon at L'Aqua on Cockle Bay Wharf was very nice, and I happily managed to arrange a dinner to catch up with a few Sydney friends with John A., Julien G., Jiri B., and Adam B. all attending. Most importantly, however - as confirmed - HPC generates an enormous social return on investment, either in cost savings or new revenue. The current figure being quoted is around $77 per $1 funded - which makes it an ultimate infrastructure investment. Of course, the most recent pandemic is a case in point, and today I chaired the researcher meeting for Research Computing Services with Tom Karagiannis and team explaining how they built a massive database of antivirals for COVID-19 related molecular dynamics simulations. It is results like these that prove the worth of supercomputing; the infrastructure that touches and improves all our lives, and yet we can rarely notice.
The Conference was especially good for getting a sense of the power of some of the truly big iron. Representatives of the two largest systems, Frontier (no 1), Aurora (no 2), and Fugaku (no 4), were present. It is humbling to see systems that have more nodes than cores in the UniMelb system (which is no laggard), and pleasing to get a good grasp of the architecture. I was quietly pleased that my own comments about HPC microprocessor trends from 18 months ago were quite accurate, and I have a special nod to the Thai supercomputing centre which has really become an impressive national facility in its own right. A nice quirk of the conference was that the voice of Siri, Karen Jacobsen was the MC for the plenary sessions. There was, of course, the opportunity to talk to many vendors who are all doing some very interesting things and making plentiful offers, although I do note that there is often a pitch in favour of monopolisation ("we'll do everything for you") rather than interoperability.
For my own part, I presented a talk the "HPC Certification Forum: An Update", which was very well received (particularly the idea of using the skill tree as a teaching structure), and the poster "HPC Training Generates HPC Results". The evening hosted by Xenon at L'Aqua on Cockle Bay Wharf was very nice, and I happily managed to arrange a dinner to catch up with a few Sydney friends with John A., Julien G., Jiri B., and Adam B. all attending. Most importantly, however - as confirmed - HPC generates an enormous social return on investment, either in cost savings or new revenue. The current figure being quoted is around $77 per $1 funded - which makes it an ultimate infrastructure investment. Of course, the most recent pandemic is a case in point, and today I chaired the researcher meeting for Research Computing Services with Tom Karagiannis and team explaining how they built a massive database of antivirals for COVID-19 related molecular dynamics simulations. It is results like these that prove the worth of supercomputing; the infrastructure that touches and improves all our lives, and yet we can rarely notice.