T he Big Mac supercomputer cluster at Virginia Tech is an amazing piece of work on several fronts, not the least of which is its bang for buck. Big Mac cost a mere five million dollars, yet it is ranked as the worldis third fastest computer. If you consider that each of the other top five fastest supercomputers cost many times what Big Mac did, you can see why the folks at Virginia Tech has caused such a stir; theyive proven that supercomputing does not have to be high priced.
Possibly taking a cue from Virginia Tech, the Australian Military has recently put together a cluster of 16 Xserves for use in that countryis war gaming effort. From a Computerworld Australia article, Apple plays 32-processor war games:
After a long stint on the sidelines of enterprise and government IT, Apple Computer has gained local traction in the defence sector. The Defence Science and Technology Organisation (DSTO) is showcasing an Xserve cluster to model battle scenarios based on the games of chess and checkers.
Dubbed Checkmate, the DSTO supercomputer-on-a-budget consists of a 16-node cluster of 1.33MHz G4 dual-processor Xserve systems (32 processors in total), each with 1G of RAM. The cluster operates on OS X 10.2.4 Jaguar, with each Xserve running multiple instances of Java-based simulation and analysis code to crunch through terabytes of what research mathematician Greg Calbert calls "second order" battle scenarios, or factors.
Second order factors are those beyond more rudimentary equations such as sheer military might and speed to consider the way scenario players interoperate and relate to each other.
"In the first order we [already] know having better tempo, better force strength [etc] to be true. We know we want better battleships, more networking - but itis a matter of how we distinguish between those factors," Calbert said.
To obtain such refinements, the G4 cluster grinds its way through specially crafted variants of what are already highly computation-intensive games.
"One of the variants of chess and checkers we play is inetwork checkersi. The pieces are connected in a network [and] we look at where they are, and exchange that [position] with other pieces [including] the values of their prospective moves. They communicate and find what is the best move.
Stop by Computerworld Australia to read the full article.