Now that the new TOP500 list is out, you might be wondering what it means in terms of the bigger picture and the global economy. But what if you could look at historic data from the TOP500 and match it against socio-economic factors such as FLOPS/capita and FLOPS/$GDP using motion charts? You might see trends that could lead to a whole new set of insights. Well, that’s exact what Berkeley’s Karl Fuerlinger has done for us HPC fans at http://top500minder.net.
I find this type of visualization is pretty well suited to interactively explore the dynamics of the HPC market. Additionally, with the recent rise of HPC in China I think it’s all the more important to look at the broader socio-economic environment of our field. Data from the 11/2011 list and the Chinese Tianhe-1A system are not included yet, but the page will be updated as soon as those results become available.
I caught up with Karl over email to find out more about the project.
insideHPC: This is an interesting site. Can you tell us a little about yourself and your day job?
Karl Fuerlinger: I’m a postdoctoral researcher at UC Berkeley and also work with people at the NERSC computing center. After a couple of years in the US I’ll me moving back to Germany to join the University of Munich and the Leibniz Supercomputing Centre in January. I’m interested in all aspects of parallel computing and work mostly on tools for performance monitoring and workload characterization.
insideHPC: What prompted you to create this site?
Karl Fuerlinger: I wanted to see what the Top500 data looks like when visualized using Gapminder-style motion charts for quite some time and finally got around doing it recently. I think the motion charts are a very intuitive technique to visualize progress over time and with the economic rise of developing countries its interesting to look at the effects it has on our field. All credit goes to the maintainers of the Top500 list and to the people who invented and implemented these nice visualization methods. I only put the two together.
insideHPC: Did you spot any surprising trends when you started to visualize this Top500 data?
Karl Fuerlinger: With the Top500 data itself I think it is interesting to see how quickly technologies and vendors come and go, with a few things such as IBM’s strong presence since 2000 that stay fairly constant. For FLOPS/capita I was hoping to see how smaller countries stack up against the dominant U.S. when you consider the country’s size. FLOPS/$GDP is interesting because someone’s got to pay for all the computing centers in the end. The orders of magnitude of these numbers are interesting too. We’re at about 60 MFLOPS per person and 1400 FLOPS per $GDP of aggregate performance in the Top500 in the U.S.
insideHPC: With the recent rise of HPC in China, why do you think it’s all the more important to look at the broader socio-economic environment of HPC?
Karl Fuerlinger: China and other emerging countries have enjoyed an amazing economic growth in the last couple of years at a time when many institutions in the western countries face cuts to education and research programs. I’m not an economist or social scientists, but these developments should be of interest to anyone engaged in education or R&D. The 2010 UNESCO science report just came out in early November and I think it clearly identifies to the growing role of knowledge for the global economy and we all know about the importance of HPC for scientific discovery. The report also shows that R&D spending as a percentage of GDP in emerging economies is up sharply over the last five years, while it stagnates or declines in the U.S. and Europe — something we should be wary about. In terms of FLOPS/$GDP Top500minder graphs show that China is on the trajectory to rapidly catching up with the U.S. and potentially overtaking it. This has happened only once before — in 2002 with Japan’s Earth Simulator.
insideHPC: Can we look forward to seeing an update when the next TOP500 list comes out in June 2011?
Karl Fuerlinger: I’ll definitely try to keep the site up to date and hopefully add some more external data sources to highlight some of the developments. Suggestions are always very welcome.