By Bob Brown, Network World -- Colorado State University's Patrick Burns could talk about grid computing 'til the cows come home. And in fact, the school's associate vice president for information and instructional technology has used grid computing to help cattle do just that. The school set up the Colorado Grid Computing Initiative (COGrid) in 2004 through $2 million-plus in grants from the Colorado Institute of Technology, plus funding and equipment from Sun, the Department of Homeland Security and others. The first project to exploit the grid was for processing animal tracking data, such as for identifying cattle in the case of a mad cow disease breakout. CSU is a land-grant institution, so is big into agricultural studies. Burns says the cattle industry is in discussions with the Department of Agriculture and others to try to develop a nationwide animal ID database, but he acknowledges that even with the power of grid, many complex questions need to be addressed regarding privacy of information and distribution to numerous sites. COGrid consists of two connected Sun 6800 servers, each with 24 CPUs, 96G bytes of memory, 1T byte of storage and linked to CSU's net at 1Gbps. Under its agreement with Sun, the school gets to keep the grid system for two years maintenance-free, then needs to work out a maintenance payment system from there. Under Phase 1 of its COGrid project, CSU got the systems in place at its data center and offered its first classes in grid computing to 80 or so students. Burns wore two hats and taught the classes in grid computing and related applications, though now he is looking forward to the school offering such courses again through a new teacher. The grid, meanwhile, is being used by more than a dozen departments at the school, including atmospheric science and mechanical engineering. Under Phase 2 of the project, a proposal for which is into the Colorado Institute of Technology, CSU would teach other institutions throughout Colorado how to use the grid, Burns says. Phase 3 would involve getting additional grid systems donated at other participating schools, so that an expanded grid could be formed, he says. What has Burns learned so far by working with the grid? "It still takes a lot of work to map an algorithm onto an advanced computing system and after that get good performance out of it," he says. "The human element is still very much required to design algorithms, implement them and test ways of doing things." Those applications requiring lots of processing power tend to do better on a grid than those that are more communications oriented, he says. ----- Copyright 2006 Network World. Supercomputing Online is a Network World Content Partner.