计算机专业外文翻译.docx
《计算机专业外文翻译.docx》由会员分享,可在线阅读,更多相关《计算机专业外文翻译.docx(34页珍藏版)》请在冰豆网上搜索。
计算机专业外文翻译
ATaleofClouds:
ParadigmComparisonsandSomeThoughtsonResearchIssues*†
LijunMei
TheUniversityofHongKong
Pokfulam,HongKong
ljmei@cs.hku.hk
Cloudcomputingisanemergingcomputingparadigm.Itaimstosharedata,calculations,andservicestransparentlyamongusersofamassivegrid.Althoughtheindustryhasstartedsellingcloud-computingproducts,researchchallengesinvariousareas,suchasUIdesign,taskdecomposition,taskdistribution,andtaskcoordination,arestillunclear.Therefore,westudythemethodstoreasonandmodelcloudcomputingasasteptowardidentifyingfundamentalresearchquestionsinthisparadigm.
Inthispaper,wecomparecloudcomputingwithservicecomputingandpervasivecomputing.Boththe
industryandresearchcommunityhaveactivelyexaminedthesethreecomputingparadigms.Wedrawaqualitativecomparisonamongthembasedontheclassicmodelof
computerarchitecture.Wefinallyevaluatethecomparison
resultsanddrawupaseriesofresearchquestionsin
cloudcomputingforfutureexploration.
Keywords:
cloudcomputing,paradigmcomparison.
1.Introduction
Cloudcomputingisaparadigmthatfocusesonsharing
dataandcomputationsoverascalablenetworkofnodes.
Examplesofsuchnodesincludeendusercomputers,data
centers,andWebServices.Wetermsuchanetworkof
nodesasacloud.Anapplicationbasedonsuchcloudsis
takenasacloudapplication.
Thisparadigmisincreasinglypopularintheindustry,
whereindustrialleaderssuchasMicrosoft[26],Google
[2],andIBM[5]stronglypromotetheparadigminrecent
years.Anearlyattempttoformulatecloudcomputing
datesbacktoatleast1997[8].However,toourbest
knowledge,theadoptionandpromotionofcloud
computinghasbeenslowuntil2007[9].
Weobservethatthehistoryofearlyindustrial
adoptionsofcloudcomputingsharesomecommon
milestoneswiththatofservicecomputing[4].For
example,ittookservicecomputing[27]alongtime(ten
yearsorso)toreceiveworldwidesupportfromleading
companieslikeIBM,Microsoft[25],BEA,andOracle.
Similarly,ithasbeenmanyyearssincetheearly
formalizationeffort[8]towardcloudcomputing.
Besides,thewideadoptionofacomputingparadigm
usuallydependshighlyonthematurityofsupporting
technologiesandindustryrecognitions.Servicecomputing
hasbecomemuchmorepopularsincethesuccessof
Webservices,althoughaWebserviceisonlyoneofthe
technologiestofulfillthenotionofserviceorientation[4].
Similarly,thedistributedcomputingcommunityhas
pointedoutthatmanydistributedcomputingtechniques
forcloudcomputinghavebeenmature[7][10][11].Many
companiessuchasDellandIBMhavebeguntoship
cloudcomputingmachines[5][10].
Lastbutnottheleast,ineitherservicecomputingor
cloudcomputing,researchdevelopmentslagbehind
industrialadoptions.Forinstance,COSCON,aleading
internationalcontainershipper,hasasuccessfuladoption
ofservicecomputing.Itsuccessfullyusedservice-oriented
architecturetoimprovethebusinessresponsibilityto
customersin2004[3].Yet,researchstudiesinserviceoriented
architecturefromthesoftwareengineering
community[19]arestillinadequate.
DespiteoursurveyovertheInternet,toourbest
knowledge,therearefewarticlestopinpointresearch
*©2008IEEE.Thismaterialispresentedtoensuretimelydissemination
ofscholarlyandtechnicalwork.Personaluseofthismaterialis
permitted.Copyrightandallrightsthereinareretainedbyauthorsorby
othercopyrightholders.Allpersonscopyingthisinformationare
expectedtoadheretothetermsandconstraintsinvokedbyeach
author’scopyright.Inmostcases,theseworksmaynotbereposted
withouttheexplicitpermissionofthecopyrightholder.Permissionto
reprint/republishthismaterialforadvertisingorpromotionalpurposes
orforcreatingnewcollectiveworksforresaleorredistributionto
serversorlists,ortoreuseanycopyrightedcomponentofthisworkin
otherworksmustbeobtainedfromtheIEEE.
†ThisresearchissupportedinpartbytheGeneralResearchFundofthe
ResearchGrantCouncilofHongKong(projectnos.111107,717308,
and717506).
‡Correspondingauthor.
2
issuesincloudcomputing.Thiswouldslowdownthe
nextresearchadvances.Wewillalleviatethisproblemin
thepresentpaper.
Inthispaper,weusetheclassiccomputerarchitecture
model[15]toprovideaqualitativecomparisonframework
tocomparecloudcomputingwithpervasivecomputing
andservicecomputing.Thequalitativecomparison
frameworkincludesthreefeatures:
input-output(I/O),
storage,andcalculation.Foreachfeature,wedrawthe
comparisonusingmultiplecharacteristics.Throughsuch
comparisons,weidentifytheconnectionsbetweencloud
computingandtheothertwocomputingparadigmsfrom
theperspectiveofsoftwareengineering.Basedonthe
connections,wedrawupafewresearchissuesanddiscuss
theminthepapertopromotefutureexploration.
Themaincontributionofthepaperistwofold:
(i)To
ourbestknowledge,weprovidethefirstqualitative
comparisononcloudcomputing,servicecomputing,and
pervasivecomputing.(ii)Wepresentaseriesofresearch
issuesincloudcomputingontopofthecomparison
framework.Theseissuespromotefutureexplorations.
Therestofthepaperisorganizedasfollows:
Section2
presentsthepreliminariesofcloudcomputing,service
computing,andpervasivecomputing.Section3introduces
ourqualitativeframeworktocomparetheabove
threecomputingparadigmsandpresentoureffortsto
identifyresearchissuesincloudcomputing.Finally,we
reviewrelatedworkinSection4anddrawaconclusionin
Section5.
2.Preliminaries
Thissectionreviewsthepreliminariesofcloud
computing,servicecomputing,andpervasivecomputing.
2.1.Cloudcomputing
AswehaveintroducedinSection1,acomputingcloud
isamassivenetworkofnodes.Thus,scalabilityshouldbe
aqualityfeatureofthecomputingcloud.Ithasatleast
twodimensions,namelyhorizontalcloudscalabilityand
verticalcloudscalability(adaptedfrom[9]).
Horizontalcloudscalabilityistheabilitytoconnect
andintegratemultiplecloudstoworkasonelogical
cloud.Forinstance,acloudprovidingcalculation
services(calculationcloud)canaccessacloud
providingstorageservices(storagecloud)tokeep
intermediateresults.Twocalculationcloudscanalso
integrateintoalargercalculationcloud.
Verticalcloudscalabilityistheabilitytoimprovethe
capacityofacloudbyenhancingindividualexisting
nodesinthecloud(suchasprovidingaserverwith
morephysicalmemory)orimprovingthebandwidth
thatconnectstwonodes.Inaddition,tomeetincreasing
marketdemand,anodecanbegraduallyupgradedfrom
asinglepowermachinetoadatacenter.
Scalabilityshouldbetransparenttousers.Forinstance,
usersmaystoretheirdatainthecloudwithouttheneedto
knowwhereitkeepsthedataorhowitaccessesthedata.
Forsimplicity,wewillrefertohorizontalandvertical
cloudscalability,respectively,ashorizontalscalability
andverticalscalabilityinthispaper.
2.2.Servicecomputing
Servicecomputing(orservice-orientedcomputing)is
anemergingparadigmtomodel,create,operate,and
managebusinessservices.Inthisparadigm,services
publishthemselvesinpublicregistries,discoverpeer
services,andbindtothelatterservicestoformservice
compositionsusingstandardizedprotocols[6].Tocreatea
servicecomposition,engineersmayuseaspecification,
suchasWS-BPEL[30],tomodelthecollaborativeneed
inworkflows.Tocarryoutindividualworkflowsteps,
softwaredevelopersmayuseWebservices,themost
popularwaytofulfillservice-orientedarchitectureinthe
industry.Asetofservice-orientedapplicationsoverthe
Webservicesthuscreatesanetworkofservices.
ServiceServiceRegistry
RegisterService
toRegistry
DiscoverService
fromaregistry
BindServiceAssociateService
Figure1.Service-orientednetwork[18].
Webrieflydescribeaservice-orientednetwork[18]to
facilitatethecomparisonintherestofthepaper.An
elementinsuchanetworkisaserviceregistry,service
consumer,orserviceprovider.Aserviceprovider
registersitselfinaserviceregistry.Aserviceconsumer
firstdiscoverstheservicefromaregistry,andthenbinds
totheservice.Aserviceprovidermayregisteritselfto
morethanoneregistry.Aregistrymayalsoassociateits
registeredservicestootherregistries,andactsasaservice
itself.Suchatreatmentonaregistryprovidesageneric
viewamongelementsinservice-orientedmodeling.
2.3.Pervasivecomputing
Pervasivecomputing(orubiquitouscomputing)[23][24]
isanotheremergingcomputingparadigm.Software(often
referredaspervasivesoftware)canbeembeddedina
constantlychangingcomputingenvironment.Therefore,
pervasivesoftwareusersdonotneedtobeconcerned
abouthowtoadjustthesoftwaretoadapttothe
surroundingcomputingenvironment.Awell-developed
environmentwillenableuserstousepervasivesoftware
everywherewithoutextraeffort.
Tounderstandandreacttoauser,applicationsuse
environmentalfeatures,knownascontexts,extensively.
Sensorscancapturethesecontexts.Toallowubiquitous
3
supporttoendusers,smartsensorsareplacedaround
userstopreservedifferentinformation,suchasthe
locations,contexts,anduser-relevantdata.
Figure2showsapervasivecomputingexample.
Sensors,mobilephonesandPDAs,desktopcomputer,and
serversareinterconnectedlogicallytoformanapplication.
SupposeanomadicuseratthetopleftcornerofFigure2
movesfromusingalapt