Grid Computing and cloud computingWord下载.docx
《Grid Computing and cloud computingWord下载.docx》由会员分享,可在线阅读,更多相关《Grid Computing and cloud computingWord下载.docx(5页珍藏版)》请在冰豆网上搜索。
3
DefinitionforCloudcomputing
∙Alarge-scaledistributedcomputingparadigmthatisdrivenbyeconomiesofscale,inwhichapoolofabstracted,virtualized,dynamically-scalable,managedcomputingpower,storage,platforms,andservicesaredeliveredondemandtoexternalcustomersovertheInternet.(IanFoster)
∙特点:
▪超大规模
▪虚拟化
▪高可扩展性
▪高可靠性
▪按需服务
▪极其廉价
▪通用性
Relation
CloudComputing
∙CloudsandGridssharealotcommonalityintheirvision,architectureandtechnology,buttheyalsodifferinvariousaspectssuchassecurity,programmingmodel,businessmodel,computemodel,datamodel,applications,andabstractions.
oThevisionisthesame
▪reducethecostofcomputing;
▪increasereliability;
▪increaseflexibility.
otheproblemsaremostlythesame
▪managelargefacilities;
▪definemethodsbywhichconsumersdiscover,requestanduseresourcesprovidedbythecentralfacilities;
▪implementtheoftenhighlyparallelcomputationsthatexecuteonthoseresources.
BusinessModel
∙Traditionalbusinessmodel
osoftwarehasbeenaone-timepaymentforunlimiteduse(usuallyon1computer)ofthesoftware.
∙Cloud-basedbusinessmodel
oacustomerwillpaytheprovideronaconsumptionbasis,suchaselectricity,gas,andwater,reliesoneconomiesofscaleinordertodrivepricesdownforusersandprofitsupforproviders.
∙ThebusinessmodelforGrids
oproject-orientedinwhichtheusersorcommunityrepresentedbythatproposalhavecertainnumberofserviceunits(i.e.CPUhours)theycanspend.
Architecture
PaaS
SaaS
IaaS
如:
SaleforceonlineCRM服务
GoogleAppEngine
AmazonEC2/S3/SQS服务
ResourceManagement
∙ComputeModel
oGrids:
▪batch-scheduledcomputemodel,don’tnativelysupportinteractiveapplications
oClouds:
▪resourcesbeingsharedbyallusersatthesametime,allowlatencysensitiveapplicationstooperatenatively
∙DataModel
oInternetComputingwillbecentralizedaroundData,CloudingComputing,aswellasClientComputing.
∙Combiningcomputeanddatamanagement
oGrid
▪Data-awareschedulersanddispersingdataclosetoprocessors
oCloud
▪data-intensiveapplicationsarenottypicalapplicationstoday,
▪asthescalesofCloudsgrow,itmayjustbeamatteroftimeformanyClouds
∙Visualization
▪donotrelyonvirtualizationasmuchasCloudsdo,eachindividualorganizationmaintainfullcontroloftheirresources
▪anindispensableingredientforalmosteveryCloud
ProgrammingModel
∙Grid
oMPI,GlobusTookit,WSRF,CoGetc
∙Cloud
oMapReduce,Sawzall,Hadoopetc
ApplicationModel
osupportmanydifferentkindsofapplications
▪HPC(highperformancecomputing)
▪HTC(highthroughputcomputing)
oscientificgateways
oinprinciplecatertoasimilarsetofapplications,exceptHPC
oadoptgatewaystoCloudresourcesforend-userinteraction
∙ThebrowserandWeb2.0technologieswillplayacentralroleonhowuserswillinteractwithGridsandCloudsinthefuture.
SecurityModel
∙Grid:
stricter
osinglesign-on,delegation,privacyetc
∙Cloud:
simpler,lesssecure
ocreateandmanageaccountinformationforend-users,andallowsuserstoresettheirpasswordsandreceivenewpasswordsviaEmailsinanunsafeandunencryptedcommunication.
onewuserscoulduseCloudsrelativelyeasilyandalmostinstantly,withacreditcardand/oremailaddress