You are on page 1of 2

TheFutureofComputerArchitecture

DavidPatterson UCBerkeley Version3:December12,20081 ComputerArchitectureisthefieldthatdesignscomputers,whichsetsthefoundationfor theentireITindustry. DespitethetremendousresourcesatlargecompaniessuchasIBMandIntel,thereisalong trackrecordofbreakthroughsfromacademicresearchersincomputerarchitecturethat haveledtonewmultibilliondollarindustries.Onereasonissimplythatmanyofthe leadingthinkersincomputerarchitectureareinacademia,asthenumberofawards,books, andpapersdocument.Anotherreasonisthatacademicresearchersarenotboundthe businessrequirementsofcompatibilitywithlegacyhardwareandsoftware.Athirdreason isthatwheninformationtechnologyischangingrapidly,decadesofexperiencemaynotbe ahugeasset,andbrightyounggraduatestudentswhodontknowwhatcantbedoneareas likelytomakeanimportantcontributionasthoseconsiderablymoresenior. Thus,inareaswherethereistremendousconsensusontheguidelinesonwhattodo, computerarchitectsatIntelandIBMusuallyleadtheway.Whenthedirectionsareless clear,academicsareoftenthepathbreakerseventhoughtheyhavefewerresources. Today,wehavehitthewallastothepracticallimittotheamountofpowerthata microprocessorchipcandissipate;inthepasteachgenerationofchipusedmorepower whilegettingmoreperformance.Weneedtoinventanewwaytogetmoreperformance withoutmorepower.Moreover,themostinterestingcomputersofthefutureareatthe extremesinscale: Thedatacenteristhecomputer.Amazon,Google,Microsoft,andothersareracingto constructbuildingswith50,000computerstorunsoftwareasaservice(SaaS). Thecellphoneisthecomputer.Millionsofcellphonesareshippedeachdayandthey areincreasinginfunctionality. Thepowerwallandtheextremesincomputersizemeanthattheoldguidelinesareout thewindow,so,ifhistoryisanyguide,wereenteringaneraofincreasingimportancefor academiccomputerarchitecture. Notethatinsuchanuncertaineratheremanychancesfornewmultibilliondollar industries,andthesenewindustriesarelikelytoflourishclosetowheretheresearchersdo
1Forthemostcurrentversionofthisessay,aswellasrelatedessays,visithttp://www.cra.org/ccc/initiatives

theirwork.WehaveseenthecenteroftheITuniverseshiftgeographicallybeforeduring erasofchange: NewYorkinthemainframeeraofthe1950sand1960s; Bostonintheminicomputereraofthe1970sandearly1980s; SiliconValleyinPCandWeberaofthelate1980stotoday. Giventhechallengesmentionedabove,othercountriesareinvestingtremendouslyinITin generalandincomputerarchitecturespecifically,sincetheyseetheopportunityinthisera ofchangetoshiftthecenteroftheITuniverseandthejobsthatcomewithit.Especially giventhecuttoacademicITresearchingeneralandtocomputerarchitecturespecifically intheUS,itsnothardtoimaginetheITcentermovingfurtherwestinthenextdecadeto BeijingorMumbai. Oneareaoftremendousopportunityincomputerarchitectureisthemanycorechallenge. Thegoalistoinventcomputersthatmakeiteasytowriteprogramsthatareefficient, portable,correct,andscaleasthenumberofcorespermicroprocessorincreasesaseasy asithasbeentowriteprogramsfortraditionalcomputers.Ifthisinventionallows softwaretousemanysimplepowerefficientcoresinsteadofasinglepowerhungrycore, thiswillresetthefoundationfortheITindustryforatleastthenext30years. Asecondopportunityisinventinganewcomputerarchitecturethatimprovescomputer securityandprivacy,problemsthatplaguetheITfield.Architectscouldremovemanyof thevulnerabilitiesoftodayscomputersiftheywerenotboundbythelegacyrequirements ofcompatibilitywithtodayscomputers.Theycouldalsoprovidenewfeaturestomakeit easiertobuildfast,secure,lowoverheadvirtualmachines,makingiteasierandsaferfor softwaretomigratebetweenthedatacenterandthecellphone. Athirdopportunityisinventcomputersthatwillremovetheperformancebottlenecks fromnew,highlyproductiveprogrammingenvironmentssuchRubyorPython.For example,theRubyonRailsenvironmentallowsprogrammerstoinventamazingnew computerapplicationsinjust1000to2000linesofcodefactorsof10to100lessthan conventionalapproaches.Exampleapplicationareasincludepersonalhealthcare, personalmemoryassistants,andpersonaldigitaleducators.Alas,RubyonRails performanceisfactorsof3to10worsethanconventionalsystems.Ifwecaninvent computersthatallownewprogrammingsystemslikeRubyorPythontoscaleupto hundredsofcoreswhilepreservingtheiramazingprogrammingproductivity,wecould unleashanewroundofexcitingapplicationsthatwillleadtonewmultibilliondollar industries,justaswevedonesomanytimesinthepast. IntelandMicrosoftrecentlyfundedtwomajoracademiccenterstotacklethesecritical challengesatUCBerkeleyandtheUniversityofIllinois.Manyotherhighlycredible proposalswerereceived,fromstrongteamspursuingdiverseapproaches.Anational initiativewouldfundfivemorecentersofexcellenceincomputersystemsandarchitecture, positioningtheUnitedStatestomaintainitspreeminenceintheITfield.

You might also like