With a bigger GDP than the United States, the European Union certainly has the economic wherewithal to field a top tier high performance computing (HPC) infrastructure. After taking a back seat to the U.S. and Japan in high end scientific computing for the past couple of decades, the Europeans now seem intent on playing in the deep end of the supercomputing pool. The renewed interested is exemplified by the Partnership for Advanced Computing in Europe (PRACE), whose mission is to build a world-class pan-European high performance computing service. The goal is to create three to five supercomputer centers — what they refer to as tier 0 centers — each housing a petascale supercomputer. The plan is to get the first systems online in the 2009/2010 timeframe.
So far 16 countries have joined up: Austria, Finland, France, Greece, Ireland, Italy, The Netherlands, Norway, Poland, Portugal, Spain, Sweden, Switzerland, Turkey and UK. (Ireland and Turkey just hopped aboard on Tuesday.) But what they need more than partners is money. The main funding instrument for PRACE is the EU’s 7th Framework Programme, which has anted up €40 million for 2008 and 2009. In the supercomputing world though, that’s just pocket change.
An article in Computerworld today talks about the European effort in getting petascale capability onto the continent, and points out some of the challenges:
The cost over 20 years of building and maintaining a petascale supercomputing facility could top €2 billion ($3.1 billion U.S.), according to Achim Bachem of Jülich Research Centre, a German HPC laboratory. He estimates annual running costs at €100 million to €200 million, with the cost of construction around twice that. Yet such systems are soon overtaken in performance by newer models, and must be replaced every two or three years to keep up with advances in technology.
Closer to home, the RoadRunner petaflop supercomputer for Los Alamos was said to cost $110 million over its three-year development and the planned “Blue Waters” multi-petaflop system for the NCSA will probably cost in excess of $200 million by the time it’s expected to be operational in 2011. The latter system will also entail building a new facility to house it.
Europe does have at least one advantage in its strong currency (€1 = $1.5), so they should be able to buy a lot more computing infrastructure for the dollar than their U.S. counterparts. And while, in general, the energy costs to feed these big iron behemoths is higher in Europe, that’s not always the case. According to the U.S. Energy Information Administration, the average cost of electricity to industrial users in the U.S. for 2005 was $0.057/kilowatt-hour. In the UK and Germany, it was $0.087 and $0.084, respectively; but in France it was only $0.050, and in Norway, a mere $0.043. Undoubtedly these numbers have changed in the past three years due to the weakened dollar (which works in Europe’s favor) and the explosion in oil prices (which works in no one’s favor).
According to the PRACE Web site, the EU’s 7th Framework Programme is “prepared to invest hundreds of millions of Euros.” Depending upon how many hundreds that turns out to be, the funding could jumpstart a reasonable petascale program in the next couple of years. But sustaining that money over several years — decades even — is probably more important. Unlike the U.S., which has enjoyed multiple supercomputing funding sources (NSF, DOE, NASA and DoD) for a generation, the European Union is just now starting to figure out how to build these kinds of super-national organizations and set priorities on a continental scale. If they can get their act together, Europe could vie for supercomputing leadership.