The feature article by Ian Foster on "The Grid: A New Infrastructure for 21st Century Science" ( Physics Today, February 2002, page 42 ) presents well the structure and great potential of the Grid. At a time when government funding tends to focus on short-term returns that directly benefit society while overlooking basic physics research, Foster has reminded us that physics remains closely associated with important new, far-reaching technological developments.
I think Physics Today missed an opportunity to deliver an important message with the publication of this fine article: Basic physics research, and basic science research in general, is often the driving force behind important developments in computing. In the European part of the world map (see Foster's Figure 4), the prominent role of CERN and of high-energy laboratories in Europe is obvious. Yet CERN (or its Large Hadron Collider) is briefly mentioned twice in the article, and only for its computing demands rather than for its contributions to the field.
During the mid-1990s, when I was the adviser to the CERN director general on member state affairs, I had to rally support for the LHC among nonscientists. The expected computing technologies resulting from the LHC and the potential for broad application of those technologies were a strong selling point. I would tell my audience that each large detector must handle more than 1015 bytes of information per year, about a million times that contained in the human genome. This fact made an impression. At the time, given the existing technology, it was impossible to handle that amount of information. The use of CD-ROMs for storage would have required a 3-kilometer-high stack of them; processing the information would have required 50 000 PCs.
But my experience has been this: Trust the physicists. By 2005, they will have found a way to meet the computing challenge, with resulting benefits for people from many walks of life. The detector collaborations, each with close to 2000 scientists, contain a vast number of highly competent people who freely exchange information and criticism. They know that, despite limited funding, they must be able to trade ideas and information and must have a successful system for that in place by the time the machine is completed.
Now, several years later, the picture has already changed. By making the best use of improving hardware and networking, scientists can reduce the storage stack by a factor of 5, and reduce the number of computers needed for processing by even more. The Grid offers the possibility of greatly enhancing the available computing power for any specific need.
Yet the capacity that LHC scientists will need has not been reached. Current technology, when fully used, already gives a factor of 10, but new developments to increase computing capacities by another factor of 10 are needed during the few years that remain before the LHC is commissioned. The Grid will play an important role in filling this need.
I think that particle physics (and heavy basic science research) as the driving force behind computing developments cannot be overemphasized. Clearly, particle physicists are not alone in demanding new and highly efficient computing means. However, carefully planned projects in the past have often fallen short of expectations, whereas those technologies that arise spontaneously out of the computing needs of physicists have paid large dividends, usually at relatively low expense.
Other arenas are motivating increases in computing power, particularly in the US, but particle physics has a specificity of its own. Very large amounts of data must be available simultaneously to a great many users. The number of physicists
working coherently on the LHC will exceed 6000, scattered worldwide. These users will need both access to information and the ability to process it. The solution to this computing challenge, once implemented, will find many other applications.
So I would like to end on a provocative note: If you want much better computing worldwide, remember to also invest in particle physics. The computing advances are likely to come faster and to be less expensive that way than through a more direct, top-down route.
Foster replies: I appreciated Maurice Jacob's thoughtful comments on the important role that physics and physicists often play in advancing information technology. I believe strongly in the use of challenging practical problems as drivers for IT R&D; such problems serve to focus on the real issues and provide rapid, if sometimes painful, feedback when apparently good ideas do not work. I am also convinced that, to achieve the order-of-magnitude performance improvements promised by Grids, we must engage not only discipline specialists but also computer scientists: for better or worse, it is no longer sufficient to view IT issues as secondary to the physics. For these reasons, I and many of my colleagues are so excited about current Grid projects, many of which involve genuine multidisciplinary partnerships focused on extremely challenging problems.
Jacob speaks more specifically to the important role that CERN has long played in IT. In a brief overview article, I could not discuss specific projects; with more space, I would have written at length about the plans and achievements of the CERN-led European Union DataGrid and DataTAG projects, the pioneering work at Italy's National Institute for Nuclear Physics (INFN), and other physics-focused Grid initiatives. (A list of project URLs can be found at http://www. mcs.anl.gov/~foster/grid-projects.) I would also have discussed the various virtual observatory projects (see Physics Today, February 2002, page 20 ), and environmental data Grid efforts. I hope that Jacob's letter and my response will clarify that physics problems and physicists are indeed central to the emergence and evolution of Grid computing.
We must all hope, as Jacob suggests, that physics will continue to have the opportunity to pose IT challenges of the magnitude associated with the LHC.
Featured Jobs
