ISC HPC Blog
Thanks to Bill
In my last blog I pointed at some of the recent developments in HPC that might turn cloud solutions into a real option for some HPC applications. Now during SC’12 at Salt Lake City Bill Kramer from NCSA published a nice article about the TOP500 which I would like to pick up to drive the discussion on HPC further.
What Bill said is all well known in the community. He points at the flaws of the TOP500 list and mentions bad practices that come with the TOP500 list. We may safely leave aside the issues that relate to a level of dishonesty which the community should not accept and which is certainly not the fault of the list but rather a lack of character of some people. The good thing about this dishonesty is that the community knows about it and has been able to deal with it. The bad thing is that the public does not know about these issues and hence was unable to judge the value of certain results.
And - I admit - I was angry about this abuse of the TOP500 list myself as our funding agencies repeatedly pointed at a well-known university that was able to reach a high ranking in the TOP500 with a low budget. And - I admit also that - I was angry when the search committee of one of the most famous universities in Europe actually asked me why my maximum ranking in the TOP500 was so low (maybe I should have explained to them the difference between Linpack and sustained application performance). The university is still famous and I am still happy I did not join them.
So, is Bill right? And are we headed in the wrong direction? And is this all the fault of the TOP500 list? The answers are difficult. I think Bill is right in mentioning the mistakes that were made over the last years when it came to making purchase decisions in HPC. I would also agree that we are headed in the wrong direction putting Linpack performance over everything else. In one of the many Exascale meetings I mentioned that our discussions are somewhat strange when we put “applications” at the end of our agenda instead of putting them first – and only after having done so derive the hardware necessities from application requirements.
I do not agree with Bill on the last issue: It is not the fault of the TOP500 list that people made bad decisions. It’s the fault of the people who made these decisions. Every center has the duty to support its users in the best possible way. Ranking high in the TOP500 is not on the list of services that has ever been asked for by users. I agree that political leaders ask for the ranking. However, every prime minister who found the way to our inauguration ceremonies in Stuttgart immediately understood when I explained in a few sentences why the TOP500 ranking is interesting but not relevant for our users. And everyone remembered that, when meeting me again for the next inauguration.
I admit, HLRS is a center that does not have to compete for the sweet grapes of TOP3 positions and hence it was always easy for me to call them sour. However, as a director of a center who has to provide services to users in research and industry it is my duty to stand up for those users. If this means to ignore the TOP500 during the procurement process, so be it.
Coming back to my original stream of thoughts: Centers will have to change. Continuing to make the mistakes that Bill has kindly pointed out will make users go away and hence in the long term will severely harm those centers which still buy systems for a high TOP500 ranking. However, those who keep improving the services for their users and keep working on workflows and applications will have a really good chance to survive. Knowing that Bill Kramer’s NCSA is a strong supporter of such a user driven approach and that countries like China, Russia, South Korea, Taiwan and Singapore follow in this trail I look forward to exciting times for HPC.
About the author
Prof. Dr. Dr. h.c. Michael M. Resch is the director of the High Performance Computing Center Stuttgart (HLRS), the director of the Department for High Performance Computing, and the director of the Information Center (IZUS) at the University of Stuttgart/Germany. Michael Resch has over 20 years of experience in HPC and was an invited plenary speaker at SC’07 in Reno/USA. He is a member of the board of the German Gauss Center for Supercomputing (GCS), and is a member of advisory councils for Microsoft, Triangle Venture Capital Group and a number of foundations. Michael Resch is a member of the advisory board of the Paderborn Center for Parallel Computing (PC2) and a member of the steering committee of the IDC HPC User Forum. He has contributed for years to the success of ISC and has been active in a variety of conferences and community activities in supercomputing over the last years. Michael Resch holds a degree in Technical Mathematics from the Technical University of Graz/Austria and a PhD in Engineering from the University of Stuttgart/Germany. In 2002 he held an Assistant Professorship at the department of computer science of the University of Houston/TX. In 2009 he was awarded an honourable doctoral degree from the National Technical University of Donezk/Ukraine. In 2011 he was awarded an honourable doctoral degree from the Russian Academy of Science.
Prof. Dr.-Ing. Dr. h.c. Dr. h.c. Michael Resch