SC09 (aka the "Supercomputing Conference") has ended, and my tired feet and aching back are now recovering, albeit slowly. I am confident that tryptophan-induced sleep from massive turkey ingestion, as part of the annual U.S. Thanksgiving ritual, will abet my recovery. But I digress …
The SC09 conference set an attendance record this year – roughly 10,000 attendees at the combined conference and tradeshow – despite the economic malaise of the technology industry and the global economy. One suspects the strong resilience of the conference may be due in part to substantial government investments in very high-performance computing (HPC). The recent announcement that a Japanese government panel recommended reducing investments in ultrascale computing may or may not be a harbinger of things to come.
The Need for Speed …
As always, a new Top500 list was revealed, and exascale computing was (literally) a hot topic, as multiple groups discussed the energy and reliability challenges inherent in building exascale systems. Jack Dongarra and collaborators summarized the results of the International Exascale Software Project (IESP), including a putative roadmap produced following the most recent workshop. A group of national laboratories and universities also announced the Hybrid Multicore Consortium (HMC), whose goal is to "address the migration of existing applications to accelerator-based systems and thereby maximize the investments in these systems."
More generally, GPUs for general purpose scientific computing were a major topic, both on the show floor, where NVIDIA's booth had standing room only presentations and in workshops and paper presentations. The real challenge of exploiting heterogeneous manycore systems lies not in achieving high performance – we know how to do that – but in creating sustainable software infrastructures and tools that embody good software engineering and maintainability principles. The hardware space is in rapid flux, as are the programming models and tools, and applications live far longer than hardware platforms.
I would be remiss if I did not point out that simplifying infrastructure and ensuring resilience and reliability at scale are some of the potential benefits of cloud computing. Clouds were also a big topic of discussion at the conference, with two cloud tutorials attracting large audiences. My colleagues and research collaborators, Rich Wolski (UCSB/Eucalyptus), Dennis Gannon (Microsoft Research) and Roger Barga (Microsoft Research) presented one of these, comparing and contrasting Amazon AWS, Eucalyptus, Windows Azure and Google App Engine as examples of infrastructure as a service (IaaS), Platform as a Service (PaaS) and Applications as a Service (AaaS).
… While Being Visual, Healthy and Green
Dona Crawford (LLNL), Wilf Pinfold (Intel and SC09 conference chair) and I recruited the panel of plenary and keynote speakers: Justin Rattner (Intel CTO), Lee Hood (Institute for Systems Biology) and former U.S. Vice President Al Gore, who gave the conference keynote. (Dona deserves full credit for landing al Gore, as she worked tirelessly to negotiate terms, processes and topics.)
Justin showed an early version of Intel's Larrabee graphics chip, itself another form of GPU, albeit with multiple cores executing the x86 instruction set. Lee highlighted the challenges and opportunities in personalized medicine, based on blood-based diagnostics. Finally, Al Gore's conference keynote wove comments about high-performance computing throughout his climate change talk. In small group conversation, I was also impressed by the former Vice President's perceptive questions about computational modeling, complexity theory and HPC applications.
As Cray/Fernbach awards chair, I was delighted to participate in the presentation of the Seymour Cray award to my friend Kenichi Miura and the Sidney Fernbach award to Roberto Car and Michele Parrinello. All three have made seminal contributions to high-performance computing. Ken talked at length about his experiences as a graduate student in the ILLIAC IV project at Illinois and how it shaped his research, reminding me of the enduring legacy of visionary projects. (See Luck is a Fickle Friend) In addition, I was part of the committee that selected my friend and long-term collaborator, Fran Berman, to receive the first Ken Kennedy award. All of this reminded me how interconnected the world of high-performance computing really is.
That's the news from Lake Wobegon, er SC09. It's time for some turkey.