The to tal co st o f the system was just 52 millio n

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Maryland will help with the research ef f o rt. Of co urse, yo u wo uldn’t want to co mpletely trust a Watso n reco mmendatio n. While Watso n was go o d eno ugh to be to urnament champ, IBM’s baby missed a f inal Jeo pardy answer o f “Chicago ” because it answered that To ro nto was a U.S. city. Source: ht t p ://www. ibm. com/p res s . One o f the f irst custo mers o f Deep Blue techno lo gies was United Airlines, which gained an ability to examine three hundred and f if ty tho usand f light path co mbinatio ns f o r its scheduling systems—a f igure well ahead o f the previo us limit o f three tho usand. Estimated savings thro ugh better yield management? Over $ 50 millio n! Finance f o und uses, to o . An early ado pter was CIBC (the Canadian Imperial Bank o f Co mmerce), o ne o f the largest banks in No rth America. Each mo rning CIBC uses a superco mputer to run its po rtf o lio thro ugh Mo nte Carlo simulatio ns that aren’t all that dif f erent f ro m the math used to simulate nuclear explo sio ns. An early ado pter o f the techno lo gy, at the time o f deplo yment, CIBC was the o nly bank that internatio nal regulato rs allo wed to calculate its o wn capital needs rather than use bo ilerplate ratio s. That cut capital o n hand by hundreds o f millio ns o f do llars, a substantial percentage o f the bank’s capital, saving millio ns a year in f unding co sts. Also no tewo rthy: the superco mputer-enabled, risk-savvy CIBC was relatively unscathed by the subprime crisis. Mo dern superco mputing is typically do ne via a technique called massively parallel pro cessing (co mputers designed with many micro pro cesso rs that wo rk to gether, simultaneo usly, to so lve pro blems). The f astest o f these superco mputers are built using hundreds o f micro pro cesso rs, all pro grammed to wo rk in uniso n as o ne big brain. While superco mputers use special electro nics and so f tware to handle the massive lo ad, the pro cesso rs themselves are o f ten o f the o f f -the-shelf variety that yo u’d f ind in a typical P C. Virginia Tech created what at the time was the wo rld’s thirdf astest superco mputer by using chips f ro m 1,100 Macinto sh co mputers lashed to gether with o f f the-shelf netwo rking co mpo nents. The to tal co st o f the system was just $ 5.2 millio n, f ar less than the typical co st f o r such burly hardware. The Air Fo rce recently issued a request-f o r-pro po sal to purchase 2,200 P layStatio n 3 systems in ho pes o f craf ting a supercheap, superpo werf ul machine using o f f -the-shelf parts. Ano ther techno lo gy, kno wn as grid comput ing, is f urther transf o rming the eco no mics o f superco mputing. With grid co mputing, f irms place special so f tware o n its existing P Cs o r servers that enables these co mputers to wo rk to gether o n a co mmo n pro blem. Large o rganizatio ns may have tho usands o f P Cs, but they’re no t necessarily being used all the time, o r at f ull capacity. With grid so f tware installed o n them, these idle devices can be marshaled to attack po rtio ns o f a co mple...
View Full Document

This document was uploaded on 01/31/2014.

Ask a homework question - tutors are online