Seti and Forgetti :)
John, here is the article:
February 28, 2001
--------------------------------------------------------------------------------
Tech Center
European DataGrid Gives Glimpse
Of Next Generation of the Internet
By BEN VICKERS
WSJ.COM
BRUSSELS -- The world's newest and biggest particle accelerator -- the Large Hadron Collider-- won't be switched on until 2005, but it's already having a profound impact on the future of the Internet.
* * *
The Hadron Collider will push the frontier of our understanding of the fundamental forces that hold subatomic particles together. In the process, it also will push computing beyond its current limits and provide a glimpse of what the next generation of the Internet could look like.
To meet the enormous computational demands of the new accelerator, the European Laboratory for Particle Physics in Switzerland, known by the French acronym CERN, is heading the development of the European DataGrid. This high-speed computer network will harness together research computers in four European countries to crunch the river of data the Collider will generate.
The goal is to complete the grid by 2004, with small tests along the way. Right now the scientists are building the network. The biggest challenge is to create what grid technologists call "middleware," the software that will manage the system.
"It can be done, we just don't know all the details yet," says David Williams, a senior scientist at CERN.
If all works out, the DataGrid will be able to handle 1,000-times more data than CERN's computer setup, plus it will allow scientists in 40 countries to access and manipulate the data simultaneously. In Europe, the new network will link 4,600 scientists at more than 3,000 institutions.
Mr. Williams sizes up the challenge the system has to meet this way: Try analyzing the data generated by six billion people, the world's population, talking down 20 telephone lines each at the same time. And the data will have to be available to thousands of users live.
The Grid, as it's called, will be able to handle much larger amounts of data than any computing system developed so far. It will have a computing power of 20 teraflops, or trillions of floating-point operations per second, compared with the three teraflops of today's most powerful supercomputers.
Though the Grid sounds as way out as the movie "The Matrix," it's a serious business. Next week in Amsterdam, representatives from software makers and other companies working on Grid technologies will join 400 scientists for the first Global Grid Forum, a week of discussions that should clarify, among other issues, standards and protocols for Grid projects.
That's a huge turnout given how new the Grid idea is. "Most scientists hadn't heard of the Grid 15 months ago," Mr. Williams says.
The Grid is a leap beyond any of the so-called distributed-computing, or Internet-computing, projects operating today, which try to amass the power of a supercomputer through networks of individual computers. One of the best known is the SETI home personal-computer network, which uses the power of idle PCs to analyze astronomical data.
Grid-like Projects
European DataGrid
www.datagrid.cnr.it
Global Grid Forum
www.gridforum.org
Computing Against Cancer
www.parabon.com/cac.jsp
Grid Physics Network
www.griphyn.org
NEESgrid (earthquakes)
www.NEESgrid.org
But all current Grid-like networks are one-way streets. "The difference between the Grid and projects like SETI, or Compute Against Cancer, is that they simply allow a research center to share the computing power of PCs, but they are not really sharing resources," Mr. Williams says. Compute Against Cancer uses PCs to analyze data on cancer patients' reaction to radiation treatment.
Existing gridlike networks allow only PC owners to lend computing power to the holder of the network, but users can't access the power of the network themselves. With the Grid, all users will have use of the giant supercomputer the pooled PCs will become.
Grid technologies will allow users to share all resources, storage capacity, computing power, data and programs -- linking PCs as if they were one computer. Users will be linked by high-speed networks, security will be handled by a single sign-on. Users won't know where the memory they are using is stored, or where their computing power is coming from.
"The tech community has realized that the next computer is going to be a world-wide computer. Everyone is talking about the post-PC era, but PCs are going to morph into an appliance," says Jim Gray, a senior researcher at Microsoft Corp.
As with the Internet's early development, a lot of questions remain about how and at what pace the Grid will develop beyond the research community. "The main limit will be human imagination," says Brian Carpenter, program director for Internet standards and technology at International Business Machines Corp. "That and your telecommunications network."
The European DataGrid initially will run as a virtual private network, over Geant, the next generation of research networks that is being built by a group of 30 European national research institutions and will be built by year's end.
It's unclear when similar network capacity will be available on a wider basis. Although the Grid can function over current telecom networks, it won't reach its full potential until heftier networks are installed and telecom rates fall.
Still, even the simplest Grid technology will be a big step forward from today's Internet experience, the researchers say, because it is so inefficient. CERN's Mr. Williams likens searching for data on the Internet to trying to fly to California without scheduled flights or reservation systems. "With the Grid, you'll just be in San Francisco." In fact, you will be everywhere at once, according to researchers.
"When the network is as fast as the computer's internal links, the machine disintegrates across the Net into a set of special-purpose appliances," says Ian Foster, one of the developers of the concept of the Grid, who works at the Argonne National Laboratory in California.
The Gridlike projects already set up give an idea of the benefits the Grid can bring. The technologies' most immediate applications could save companies billions of dollars in computing costs. Intel Corp. already uses its in-house computers in a network for supercomputing work, and estimates it has saved $500 million (548.1 million euros) over the past 10 years.
One of the main concerns of the science labs right now is ensuring the debates about Grid technologies are as global as possible and that everyone agrees on one set of standards and protocols.
The balance is tipping in favor of a set of tools based on the open-code Globus Toolkit, which was developed by U.S. researchers at the universities of Chicago and Southern California. The U.S. National Aeronautics and Space Administration is using Globus as the basis for its Information Power Grid, which will link research institutions in the U.S. The European developers have chosen Globus, as well.
But already there's a snag. "Our colleagues in Italy have found that the Globus tools sometimes become unstable when the system is heavily loaded, so we have to stabilize them. That will be the first phase of our work," says Fabrizio Gagliardi, head of the Grid project at CERN.
The large computer firms are collaborating with scientific researchers but are keeping their options open. Microsoft is working with Globus "to ensure people in a Windows environment can use Globus," says Microsoft's Mr. Gray. But Microsoft and Sun Microsysems Inc., for example, recently unveiled competing Internet-computing strategies.
Others play down the risk of a battle over standards. "We want to collaborate with those who are setting this up," says Mike Turnill, principal product manager in Europe for Oracle Corp., which is working closely with the science teams developing the Grid concept. "We are not in a position to go about proposing our own standards."
IBM, which is funding the Grid meeting in Amsterdam, expects one standard to become universal for Grid technology. "The more open a standard is, the better chance it has of dominating," says Mr. Carpenter.
Grid applications also are being evaluated by the Internet2 consortium of 180 U.S. universities that is working on improving Internet networks in the U.S. Internet2, which groups a diverse range of projects all with the goal of improving the Internet, is linking with companies and research centers to look at technologies that will improve Internet efficiency.
The Grid "has a truly international dimension and will work well with the investments we have made into Linux," says Caroline Isaac, high-performance-computing manager at IBM in the U.K. IBM announced a $1 billion investment in the Linux open-source operating system in December to push the use of Linux in the commercial field.
The first experiments in Grid-type networks date from the mid-1990s. The concept of a global Grid was developed in the book "The Grid," by Ian Foster of the U.S.'s Argonne National Laboratory in Chicago and Karl Kesselman of the University of Southern California, which appeared at the end of 1998.
The Gridlike networks now running, mostly in the U.S., so far have limited capabilities. One of the best-known is SETI home, which links private computers in a network that analyzes data from the U.S.'s National Astronomy and Ionospheric Center telescope in Puerto Rico in search of signs of extraterrestrial intelligence. Private computers carry out calculations for the coordinating organization while their owners aren't using them.
The SETI home project, which cost the organizers $500,000 to put together, has greater computing power than IBM's $100 million-plus White ASCI computer. The SETI, or Search for Extraterrestrial Intelligence, project has nearly three million home PCs onboard. Their owners have donated their unused computer time to the project.
Taking that idea a step further, Juno Online Services Inc., a large U.S. Internet-service provider, announced recently that it would seek to harness the idle computing capacity of users of its free services. Juno hopes to sell the supercomputing power that its millions of subscribers could provide to pharmaceutical firms for research and drug modeling.
Juno's announcement sparked some concerns about privacy and security, and the issue is likely to resurface with the Grid. But developers are confident Grid technologies can be secured to ensure against unauthorized access or use.
Something else may be worrying CERN's scientists. "Europe has a great tradition of starting out in front and doing its best to end up second," says Mr. Gagliardi, the head of CERN. "I hope that with the Grid we can show that we can do things just as well as others, or better." |
