Addison Snell started the debate with the theme of the international expanding of HPC. Is it important where a supercomputer is located, he asked the panelists.
Technically it is not important where the supercomputer is hosted, according to Michael Resch, but location does matter where the codes are involved. It is important to know where the codes are designed. From an economical point of view, it depends on the type of country.
Satoshi Matsuoka pointed out that energy matters as to where the machine is located. He also thought that putting people together is essential.
Andrew Jones agreed that for science it does matter where the supercomputer is hosted, also because of access policies and local skills base.
Jean-Marc Denis emphasized that location does matter a lot in order to safeguard the control of the supercomputer technology. It is also importnat for the economic growth in a country.
Addison Snell asked whether the focal point would change. Jean-Marc Denis confirmed the rising of China and Japan. Satoshi Matsuoka added that we should not forget Korea, India, and Russia for environment, health, and other applications.
Michael Resch stated that Russia is very interesting and prestigious. He had some questions about the sustainability but definitely put Russia on the agenda.
Addison Snell changed the topic to Cloud computing. This technology is getting adopted for HPC applications.
Andrew Jones very positively saw the engineer at the desktop. Jean-Marc Denis noted that it is a good solution for the commercial world first since they do not have HPC. It will work for commercial companies, for a limited amount of time. Satoshi Matsuoka described cloud as the low end of supercomputng to some extent. Over time and very likely, the industry will pick up supercomputing, he predicted. Then the cloud solution might be more appropriate for the high end. Michael Resch was very bold and honest: "I am a researcher. The term cloud has not changed anything in terms of research. And if you ask industry about cloud, the industry answers: so what? The cloud will go away, I do not care about it."
Addison Snell remarked that private and public cloud are like American and European football: there is some convergence but there are definitely some important differences in implementation.
Jean-Marc Denis stated that the cloud is an HPC system and Satoshi Matsuoka added that cloud is also about virtual machines.
Coming to exascale, Satoshi Matsuoka noted that a lot of topics are not solvable if we have not exascale systems. Michael Resch added that it is the next logical step. We should ask ourselves: what kind of problems can we solve with exascale? This is a huge challenge.
Andrew Jones had a big yes for exascale but exascale is not the only driver in HPC. You need both exascale and terascale systems. Michael Resch ended this topic with the philosophical but essential question for the researcher: is our work useful for the general community?
Addison Snell then went over to the relevance of the TOP500 as the clock was ticking.
Jean-Marc Denis thought that it is not the list which is important but the ecosystem and according to Andrew Jones the list includes lots of useful information. Satoshi Matsuoka agreed in saying that one should not overestimate the list but one should neither underestimate it. Michael Resch played the devil in disguise noting that such list maybe statistically can be relevant but otherwise not.
GPU was the next item on Addison Snell's list but Michael Resch blocked it off with a simple no. Satoshi Matsuoka would rather speak of many-core, be it CPU or GPU. Andrew Jones agreed that GPU is not the final solution and Jean-Marc Denis described GPU as a technology-enabler for exascale computing.
Then Addison Snell summed up some segments in application and asked his panel whether it is HPC or not.
To online games Andrew Jones said yes but Michael Resch no. Animation rendering received a no from Jean-Marc Denis as did financial risk management from Micael Resch. Also a no for computer engineering on a dual socket from Satoshi Matsuoka but a yes to Google from Andrew Jones, Jean-Marc Denis and Satoshi Matsuoka. Should it surprise us that Michael Resch answered no?
Addison Snell also wanted to know which aspect of ISC'11 had been a surprise for the panel and which aspect was more of a disappointment?
For Jean-Marc Denis the whole event was clearly the winner whereas the absent people were definitely the losers. Andrew Jones marked the K computer as the big winner as did Satoshi Matsuoka and Michael Resch. For Andrew Jones the loser was GPU. Satoshi Matsuoka was disappointed about the USA machines that had lost in the TOP500 because these machines can be very competitive. Michael Resch marked Hamburg as a loser since the ISC conference will leave the city in 2013.
Addison Snell ended by thanking the panelists and the audience for their wonderful co-operation in this Formula 1 analyst crossfire discussion.