SEARCH:

advertisement


The Grid: A Computer Web for Astrophysics and More
By Katherine Freese
Special to SPACE.com
posted: 12:08 pm ET
28 September 2000

grid_000927

Forget about The Matrix. It's time for the Grid -- the next wave in computer science as driven by the world's most ambitious particle and astrophysics experiments.

The Grid has the potential to manage some collossal physics experiments, not to mention advance communications to the virtual potential first imagined decades ago.
   Images

A computer reconstructed image of an event at the particle accelerator in CERN, Geneva, where electrons and their antiparticles are smashed into one another. Future experiments of this type will generate huge amounts of data that can only be handled by a

A production Grid showing the integration of data generation facilities, storage, computing and networks, plus tools for scheduling, management and security.
   More Stories

Intelligence in the Cosmos: Flesh or Machine?


Data Glut Threatens Space Archivists


Supercomputers Foretell Cosmological Findings


NSA to Outsource Non-Classified Information Technology

   Related Links

The GriPhyn Project


CERN

In late 1990, Tim Berners-Lee, a scientist at CERN, the European Laboratory for Particle Physics near Geneva, Switzerland, invented the World Wide Web. Now CERN experiments are pushing for a mammoth expansion of computing capability. Here in the United States, the National Science Foundation has announced a grant of $11.9 million dollars toward developing the Grid Physics Network, or GriPhyN (pronounced "griffin") at universities around the country.

Two experiments at CERN, the largest high-energy particle physics facility in the world, are expected to start taking data in 2005. These experiments will accelerate particles up to the fastest energies ever produced in a laboratory.

One of the goals is to find the elusive explanation for why particles have mass; why does the proton weigh 2 times 10^-24 grams and why does the electron weigh 2,000 times less? These experiments, called ATLAS and CMS, will each employ roughly 2,000 physicists, the largest number ever to work on one experiment.

"These experiments will start to churn out more data than humankind has ever seen," said Homer Neal of the University of Michigan, a member of the ATLAS team. "They're going to produce petabytes of data per year. Right now we don't know how to handle that much data."

Petabytes of data means a thousand trillion bytes. This is the amount of data that can be stored by a million personal computer hard drives. These data are produced in Geneva, but scientists in universities and laboratories all over the world will want to use and analyze it.

Here's where the Grid becomes essential. Neal likens the situation to sitting in Michigan and trying to sip from a soda at the cafeteria in Geneva: you'd have to suck through a 3,000-mile- (4,800-kilometer-) long straw. That's why we'll need the Grid -- a network of computers all over the world that will work together to deal with these huge databases.

No one computer is powerful enough to do the job, so a worldwide pool of computers will be necessary.

"We'll collect resources of universities, supercomputers, national labs, etc. to build one great virtual computer," said Paul Avery, lead scientist of the GriPhyN project and University of Florida physics professor. The Grid will know which computers in the pool are sitting idle at any given time; when a scientist submits a job, it will be sent out to whatever computer isn't being used.

For example, even for a scientist in Michigan, maybe it will turn out to be most convenient for the Grid to send both the analysis program written by the scientist and the relevant part of the data to the University of Chicago; there the data gets analyzed and the results sent back to the scientist in Michigan.

"GriPhyN could be thought of as a Napster for scientists, where the tunes being downloaded are not purloined hits but crucial insights into the nature of the universe," said project co-leader Ian Foster, professor in Computer Science at the University of Chicago.

The Grid gets its name from its similarity to the network of power grids that bring us our electricity.

When we plug an appliance into the 110-volt socket in the wall, we don't know who generated the power, we don't know what plant the power came from and we don't know how it got here. Whether it comes from a hydroelectric power plant in Colorado or from somewhere in Idaho, we expect the lights to turn on and the refrigerator to stay cool.

In the Grid, the same idea is being applied to computer interconnectivity. When scientists submit a processing job to this worldwide network of computers, the only thing they care about is that the job gets done. They don't know which machine (or machines) the work gets farmed out to. They don't care whether it's the National Computer Center in Indiana or some PCs in Arizona that are doing the job, as long as it's completed. The average household PC has the computing power of a $200,000 machine in the 1950's. The Grid is going to network together machines from all over the world.

One day the Grid may be commonplace, with every household plugging in. Currently the GriPhyN project includes the two experiments at CERN; The Sloan Digital Sky Survey, which will gather information about hundreds of millions of galaxies and the Laser Interferometer Gravitational-Wave Observatory, a gravity wave detector.

Future applications for the Grid are mind-boggling. A medical school professor could use the Grid to present a holographic representation of a cadaver, with the images loaded with data representing millimeter-thick slides of tissue such that students could conduct a virtual dissection. The Grid also could manage the quantity of data needed to create avatars to simulate co-present conversations and interactions.

In the near future scientific endeavors such as the Human Genome Project and the Earth Observing System will need to handle huge databases. The growth of computer power averages about 1 percent per month. The improvement of cost/performance over the last 35 years has been on the order of 100 million. Imagine where we'll be a hundred years from now.

Katherine Freese is a professor of physics at the University of Michigan.


     about us | FREE Email Newsletter | message boards | register at SPACE.com | contact us | advertise | terms of service | privacy statement      DMCA/Copyright

     © Imaginova Corp. All rights reserved.