If the Icelandic volcano gods permit, this year’s International Supercomputing Conference in Hamburg, Germany, will be the best-attended and the most exhibitor-laden show in the event’s 25-year history. The ISC organizers expect upwards of 2,000 attendees and around 150 exhibitors, both of which would be records.
The conference starts Sunday, May 30, and runs through Thursday, June 4, giving the HPC faithful 5 days of unrelenting supercomputing revelry. I’ll be there from start to finish, endeavoring to bring you the highlights of this year’s event with our special live coverage from HPCwire. And there should be plenty to cover.
As always, the big problem for us in the HPC journalistic biz at these big supercomputing shows is finding the gold nuggets amidst a lot of shiny-looking news, sessions, and exhibits. In this article, I’ll attempt to point out what I consider that can’t-miss happenings at this year’s conference.
One of the traditional events at ISC is the reshuffling of the TOP500 list, which represents the top supercomputing systems in the world by Linpack prowess. As far as what to expect this time around, I covered most of this in Wednesday’s blog entry. In that post, I surmised that the top systems would only see one or two new petaflop entries, but since then, I found out about two other possible candidates.
One is the new 1.25 (peak) petaflop Tera 100 system from Bull, which was installed for the French Atomic Energy Authority (the CEA). According to the Thursday press release, the machine was just powered up on 26 May, so presumably they missed the deadline earlier this month to turn in the Linpack benchmark results for the June list (although perhaps they did some lab benchmarking before deployment).
The second system won’t be officially announced until the week of ISC, so I can’t really say anything about it yet, except that it too is a petaflopper and it’s a brand new machine. If the owners got their Linpack results in on time, it will almost certainly be a top 5 system.
Turning to the conference proper, I’d like to point to a couple of keynotes that I think will be of interest to everyone. The first is Kirk Kaugen’s opening day keynote on Monday. Kaugen is vice president of the Intel Architecture Group and general manager of the company’s Data Center Group. He’s supposed to talk about scale-up and scale-out technology as it applies to HPC, but according to a recent Intel blog post, Kaugen will also talk about how they’re going to steer their Larrabee processor technology into the high performance computing realm. In December 2009, Intel revealed it had ditched Larrabee for the discrete high-end graphics market but left the door open to using the manycore technology for “throughput computing,” so this refocus on HPC is not too big of a surprise.
The second keynote that should not be missed is Thomas Sterling’s look back on the year in high performance computing. Something of an ISC tradition, Sterling always manages to make his year-in-review talk entertaining and informative.
On Tuesday and Wednesday are the two Hot Seat sessions, where execs from some of the big HPC vendors are scrutinized by a panel of “inquisitors.” A partial list of companies participating in the event includes: Bull, NEC, IBM, Cray, Microsoft, HP, and Fujitsu. A surprise entry is Oracle, who will be represented by Sun Microsystems alum, Marc Hamilton, who now has the title of Vice President, HPC Sales Support. If I were an inquisitor, I think my first question would be: “What are you doing here and what have you done with my Sun HPC servers?” I’m guessing someone will ask that question, albeit more tactfully.
HPC analysts will be working the conference pretty hard this year. John Barr from the 451 Group will be presenting HPC market trends and forecasts as part of the HPC Advisory Council European Workshop on Sunday afternoon. And on Monday, IDC will host its traditional breakfast analyst briefing, where the analysts will provide their own take on 2010 trends and deliver some predictions for the year ahead. And finally, Addison Snell, who heads InterSect360 Research, (along with yours truly) will be providing some real-time analysis and commentary at the conference with our ISC podcast series on Tuesday and Thursday.
As usual, all of the major HPC vendors will be exhibiting at ISC this year, including a gaggle of smaller Europe-based HPC companies that don’t usually make it to the larger US-based supercomputing conference in November. As I mentioned above, Oracle will be attending this year, and this will represent the first instance of the organization at an HPC venue. Up until now the company has said precious little about its HPC aspirations, so it will be interesting to see how they’re positioning themselves in this market after the Sun merger.
NVIDIA will also be attending this year (as a co-exhibitor with Microsoft), representing the first time the GPU maker has made the trek to ISC. With GPU computing storming into the HPC landscape this year, it’s little wonder that NVIDIA wants to get in on the fun. I counted 11 GPU computing presentations at the conference, and I’m guessing we’ll see some buzz about the new top-end supers equipped with (or soon to be equipped with) the latest Fermi hardware.
There are a boatload of worthwhile presentations at the show that are too numerous to list. But I’ll mention a few that look particularly interesting and you can follow the links to get the details:
- Parallel Computing in the Years to Come
- High Performance Computational Life Sciences – The Challenge for HPC Systems
- Current HPC Projects in Europe, Japan & USA
- How Will Supercomputer Centers Look Like in the Future? (Panel)
- High Performance Cloud Computing
- International Exascale Software Project
- New Memory & Storage Hierarchies for HPC – Opportunities & Challenges (Panel)
Of course, there are social events in the evening, but if you happen to catch me at any of them, just remind me I’m supposed to be working. Otherwise, I hope to see you there.