|
The Grid
GRID Computing: the large-scale integration of computer systems (via high-speed networks) to provide on-demand access to data-crunching capabilities and functions not available to one individual or group of machines.
What is the Challenge? (The Need)
"By linking digital processors, storage systems and software on a global scale, grid technology is poised to transform computing from an individual and corporate activity into a general utility."
-- Ian Foster (http://www.globus.org), "The GRID: Computing without Bounds", Scientific American "We see grids as the next big thing in computing and the evolution of the Internet."
-- Irving Wladawsky-Berger, Vice-President, IBM
The integration of high-performance computers and high-speed networking promises to revolutionize computational problem solving through turning computing into a general utility. Ian Foster, of the Globus Project, comments, "We would not accept a situation in which every home and business had to operate its own power plant, library, printing press, and water reservoir. Why should we do so for computers?" (http://www.sciam.com) Just as the electricity "grid" allows us to pull as much power as we want without needing additional generators in the shed, new network protocols operating across high-speed networks promise to end the era of self-contained computing.
Already, clusters of computers working together over a high-speed network are replacing even the gruntiest supercomputers for a fraction of the cost. New Zealand's own Massey University has tested this hypothesis by creating a virtual supercomputer (named Helix) by linking 66 dual Athlon processors via a high speed network which can compute at a rate of 230 gigaflops per second, making it by far the fastest "machine" in the country (and no. 304 in the world - www.top500.org). Its nearest rival is NIWA's Cray computer, which can compute about 115 gigaflops..
Most of Helix's work will be bioinformatics applications to support research from the Allan Wilson Centre in Auckland, but it will also be used by Massey mathematics staff to model ice flows in the Antarctic; in collaboration with Auckland University to model biological processes; and by quantum physicists. There may also be interest from biomathematicians, and the supercomputer could even draw the attention of the movie industry for their special effects calculations..
Chris Messon, who heads Massey University's parallel computing research centre, commented in March that he would like to see Helix connected to other clusters of computers and mainstream supercomputers. "A grid would be available to researchers to use around the country," says Messon. "Over time we're expecting small supercomputers, either like or slightly different to Beowulf clusters, to join." (http://www.computerworld.co.nz). Messon says the main factor holding the project back is the lack of a high-speed network around the country.
How Can NGI Help? (The Use).
The amount of data collected today is increasing rapidly. This is especially true for any organisation working in research and development, be it designing cars or designing genes. Currently, Gr technology is being used by General Motors and Ford Motor Co. to analyse vehicle crash tests; Adventis, GlaxoSmithKline, and Pfizer to evaluate human disease and develop treatments; and Pratt & Whitney to test computer-aided simulation of engines. And with the power of GRID technology, simulations can go into even more detail, providing better insight and creating better products at substantially lower costs. However, computational ability aside, GRIDs require connectivity. And that's where NGI comes in..
For example, the amount of data collected by Monsanto scientists on animal and plant genes fills CDs that would stack 500 feet (152 metres) high - enough to send even the most powerful supercomputer into cardiac arrest and choke the fastest of today's networks. Relying on supercomputers, a gene-analysis job could take as much as six weeks to finish, thus limiting their analysis to 10 to 50 genes per year (http://www.gridtoday.com/02/0617/100006.html). Utilising GRID technology across a high-speed connection, a gene-analysis job today takes less than a day, increasing the analysis capabilities fifty-fold over what was possible five years ago..
However, even Massey's Helix system cannot handle some very large problems, and the logical strategy is to cross-link it with other systems. To do this would require something in the order of 10Gbps Ethernet technology with extremely low latency (delay of the signal) (http://www.computerworld.co.nz)..
To get an idea of how fast "fast" can be, the Oak Ridge National Laboratory (in the US) has recently put in a computer link to Atlanta, Georgia, which is 200,000 times faster than the fastest dial-up connections typical of home computers. Such a connection could download a file the size of the film "Gone With the Wind" in 6 seconds (http://www.gridtoday.com/02/0826/100290.html). These Gigabit-speed connections take away constraints that limit significant advances in science and economic development, and as a side benefit, facilitate GRID technologies that will dramatically decrease the costs and time associated with "grunty" computations. The bottom line is decreased cost and increased capability, which opens the doors for commercially relevant research and development.
The Bottom Line (The Value).
NGI combined with GRID computing protocols could one day transform the economics of computing. For example, a supercomputer can cost US$30 million for a middle-of-the-road machine. But by aggregating the power of smaller, more affordable computers, GRIDs can outmuscle the largest supercomputers for a fraction of the price - around US$25,000 for the basic software installations (http://www.gridtoday.com/02/0617/100006.html). (In comparison, Massey's Helix system had a price tag of NZ$250,000 when the hardware costs were included.).
This potential thousand-fold decrease in cost means that supercomputer-style computational power is within reach the reach of many of New Zealand's research organisations and businesses, thus removing yet another costly constraint to global competitiveness. By connecting existing clusters and supercomputers, the goal of a "computational utility company" is not that far away. With NGI infrastructure and GRID technology in its toolbox, it will be the extent of Kiwi ingenuity, not cost, that will determine New Zealand's future.
|
|
