Seamless methods of sharing resources and cheap high-performance computer power are among the promises that grid technology has been offering for quite some time now. Although the notion of a massive virtual network has been around for several decades in computer science circles, a clear concept of how the shared power could work only emerged in the late 1990s. More recently, grid computing entered centre stage; its potential impact on how both research and business will be conducted in the future has been has been compared with that of the Internet.
It seems like the perfect time for early career researchers to get their teeth into the technology. Physics-trained, Oxana Smirnova more or less sidestepped into grid computing because she needed it to move her research forward. She studied nuclear physics at Moscow State University to the M.Sc. level in her native Russia and in 1994 went to the Particle Physics Department at Lund University in Sweden to do her Ph.D. project.
As far back as 1992, she was a regular visitor at the European Organisation for Nuclear Research ( CERN) in Switzerland and working on DELPHI (a particle physics experiment studying electron-positron collisions). Before her postgraduate studies, she enjoyed CERN computing power for DELPHI analysis. But it wasn't until her postdoc in 2001 that she got to the stage that she "simply didn't have enough computer power to continue our research analysis" (on particle physics), so she started working directly on grid projects.
For researchers like Smirnova who need high-performance computers with an extremely high processing speed, massive storage capacity, and an ability to transfer large amounts of data, grid fits the bill perfectly. Networking up to 1000 processors is definitely a case of strength in numbers. Forget gigabyte; grid technology ventures into the realm of petabyte (1015) storage capacity. In hardware terms, it is equally appealing, as your everyday PC (as part of a cluster) can be part of a grid, overriding the need to have conventional and expensive supercomputers.
As a research assistant (a position in Sweden that is roughly equivalent to that of a U.S. assistant professor) at Lund University, Smirnova is now working full time on grid projects, developing grid architecture for the particle physics detector, ATLAS. But she hasn't turned her back on physics and in fact spends a fair share of her time working side by side with physicists in CERN. Having firsthand knowledge of what the grid "consumer" needs is an advantage but not a must. In fact, CERN also employs quite a number of computer scientists. "Although physicists were pushing for the solutions, the whole idea started in computer science," admits Smirnova. She also sees the sector attracting researchers' interests in a wide variety of disciplines ?? all needing supercomputing power.
Finnish computer scientist Arto Teräs, who has undertaken summer projects at CERN, agrees. "Physics was certainly one of the first applications but researchers in fields as diverse as space to biosciences are also beginning to take a closer look at grid." Teräs has an M.Sc. in computer science from Helsinki University of Technology and is currently involved in building a grid system (M-grid) based on Linux PC clusters at CSC, the Finnish IT Center for Science.
CSC is a nonprofit company owned by the Finnish Ministry of Education, the aim of which is to provide IT computational support to universities and research institutes. Therefore, Teräs's position is not in research but is focused on the operational level between the servers and end users that he defines as "middleware." "Researchers are our customers," he explains. "Our task is to find solutions to make grids feasible, be it software ones to issues of user administration, putting these [solutions] into production, and then helping the researchers to use them."
Indeed, it appears that the focus in grid computing right now is at this operational level. Smirnova see the greatest challenges for those working in grids at the administrative and political levels, rather than the technical one. There is no precedent. "No one has a clear concept on how computing power will be shared between organisations and between countries," she says. She sees the trend in funding reflecting this. "Currently, there isn't much money going into grid R&D, but funding is more likely to be assigned for operational work, user support etc.," she remarks.
Don't get hung up on the name
So what about those hungry to get to grips with fundamental grid research? Although Teräs has a personal interest in core grid technology, he believes the most exciting part is yet to come. "What people will build with grid[s] in the future, the [research] applications will be the most fascinating aspect." Smirnova also sees lots of opportunities "for eager creative young researchers." Teräs suggests that those interested in studying or research about the grid should not get hung up on the name. "I think that the best place to do grid-related research is a good computer science department or any science department that has an interest in computer networks and collaborative projects, it's not necessary to find a prelabelled 'Grid Lab.' "
He gives an example of a new project that has two Finnish partners, CSC and the Department of Astronomy at the University of Helsinki: the development of a distributed data analysis system for extensive astronomical data. The project is being led by the European Southern Observatory and co-ordinated by CSC. Teräs explains, "This is a good example of a project aiming to help scientists to access data scattered around the world using grid technologies. There will be similar projects for different fields of science emerging, giving opportunities for both computer scientists and people competent in that particular field of science."
Teräs envisages grids evolving on such a disciplinary basis and need. When the technology matures, he predicts, the commercial application will also come to light. Smirnova also sees a very healthy future for grid technology. Although she herself is more than happy in academia, "If it proves to work well, there will be a lot of interest from industry." For businesses that handle large amounts of data, switching to grid-based computing services may simply be a smarter investment than upgrading their own hardware.
Both Teräs and Smirnova believe grid technologists will need to have good negotiation, diplomatic, and people skills. Smirnova views security issues that come along with grid computing one of the most challenging. Teräs agrees, "In order to get different organisations with very different approaches to share resources and at the end of the day trust each other isn't going to be easy." Being convinced of the potential fruits of collaboration certainly helps, but enjoying dealing with so many people won't be for naught either. Teräs sums it up with, "The social part of the job is very important. I've met some very nice people and have had a lot of fun."
Editor's note: If you are interested in reading more about Grid technology, take a look at Surendra Reddy's piece. Reddy gives an overview of this field, which is predicted to have a major influence on how researchers conduct their work in the future.