intercloud-global.pdf-INTERCHANGE OF GLOBAL RESOURCES

INTERCHANGE OF GLOBAL RESOURCES

1. UNIT-3

2. INTER-CLOUD RESOURCE MANAGEMENT
ExtendedCloudComputingServices(thevariouscloudservicemodelsandtheirextensions)
•Figureshowssixlayersofcloudservices,rangingfromhardware,network,andcollocationtoinfrastructure,
platform,andsoftwareapplications.
•ThetopthreeservicelayersasSaaS,PaaS,andIaaS.
•ThecloudplatformprovidesPaaS,whichsitsontopoftheIaaSinfrastructure.ThetoplayeroffersSaaS.
Althoughthethreebasicmodelsaredissimilarinusage,theyarebuiltoneontopofanother.

3. INTER-CLOUD RESOURCE MANAGEMENT
ExtendedCloudComputingServices(thevariouscloudservicemodelsandtheirextensions)
•Thebottomthreelayersaremorerelatedtophysicalrequirements.ThebottommostlayerprovidesHardware
asaService(HaaS).
•Thenextlayerisforinterconnectingallthehardwarecomponents,andissimplycalledNetworkasaService
(NaaS).(allowingcompaniestosetuptheirownnetworksentirelywithouthardware).VirtualLANsfallwithinthescopeofNaaS.
•ThenextlayerupoffersLocationasaService(LaaS),whichprovidesacollocationservicetohouse,power,and
secureallthephysicalhardwareandnetworkresources.(LaaSisthefacilitythatoffersspacewiththeproperpower,coolingand
securitytohostbusinesses’computinghardwareandservers).
•ThecloudinfrastructurelayercanbefurthersubdividedasDataasaService(DaaS)andCommunicationasa
Service(CaaS)inadditiontocompute.

4. INTER-CLOUD RESOURCE MANAGEMENT
ExtendedCloudComputingServices(thevariouscloudservicemodelsandtheirextensions)
•AsshowninTable,cloudplayersaredividedintothreeclasses:
➢CloudserviceprovidersandITAdministrators
➢SoftwaredevelopersorVendors
➢EndUsersorBusinessUsers
•ThesecloudplayersvaryintheirrolesundertheIaaS,PaaS,andSaaSmodels.
•Thetableentriesdistinguishthethreecloudmodelsasviewedbydifferentplayers.(Tableshowshowthreeplayersview
thethreecloudmodels)
•Fromthesoftwarevendors’perspective,applicationperformanceonagivencloudplatformismost
important.(Designingapplicationwithoptimizedperformancewrttime,space,worksforallscenario)
•Fromtheproviders’perspective,cloudinfrastructureperformanceistheprimaryconcern.(optimizedCPUutilization,
Storageutilization)
•Fromtheendusers’perspective,thequalityofservices,includingsecurity,isthemostimportant

5. INTER-CLOUD RESOURCE MANAGEMENT
Extended Cloud Computing Services
1.Cloud Service Tasks and Trends
•ThetoplayerinthecloudserviceisSaaSapplicationsforbusinessapplications.
•Forexample,CRMisheavilypracticedinbusinesspromotion,directsales,andmarketingservices.
•CRMofferedthefirstSaaSonthecloudsuccessfully.
•Theapproachistowidenmarketcoveragebyinvestigatingcustomerbehaviorsandrevealingopportunitiesby
statisticalanalysis.
•SaaStoolsalsoapplytodistributedcollaboration(Googledocs),andfinancialandhumanresourcesmanagement.
•Thesecloudserviceshavebeengrowingrapidlyinrecentyears.
•PaaSisprovidedbyGoogle,Salesforce.com,andFacebook(Facebookservice),amongothers.
•IaaSisprovidedbyAmazon,WindowsAzure,andRackRack,amongothers.
•Collocationservicesrequiremultiplecloudproviderstoworktogethertosupportsupplychainsinmanufacturing.
•NetworkcloudservicesprovidecommunicationssuchasthosebyAT&T,Qwest,andAboveNet.

6. INTER-CLOUD RESOURCE MANAGEMENT
Extended Cloud Computing Services
2. Software Stack for Cloud Computing
•Theoverallsoftwarestacksarebuiltfromscratchtomeetrigorousgoals.
•Developershavetoconsiderhowtodesignthesystemtomeetcriticalrequirementssuchashighthroughput,
HA,andfaulttolerance.(Developersneedtothinkindesigningasoftwarethattakescareofalltheselayers/servicesatdifferentlevels
andprovidingtheseservicestousersinsuchawaythemeetrequiredthroughput,HA,andfaulttolerance).
•Eventheoperatingsystemmightbemodifiedtomeetthespecialrequirementofclouddataprocessing.(EvenOS
thatrunsatclouddatacenterneedtobemodifiedthattakescareofalltheseservicesNaaS,LaaSetc.)
•theoverallsoftwarestackstructureofcloudcomputingsoftwarecanbeviewedaslayers.Eachlayerhasits
ownpurposeandprovidestheinterfacefortheupperlayersjustasthetraditionalsoftwarestackdoes.
However,thelowerlayersarenotcompletelytransparenttotheupperlayers.(SaaSlayertakescareofproviding/sharing
softwarestocustomers,PaaSSaaSlayertakescareofprovidingplatformtocustomers,IaaSlayertakescareofsharinghardwarestocustomers)

7. INTER-CLOUD RESOURCE MANAGEMENT
Extended Cloud Computing Services
3. Runtime Support Services
•Asinaclusterenvironment,therearealsosomeruntimesupportingservicesinthecloudcomputing
environment.
•Clustermonitoringisusedtocollecttheruntimestatusoftheentirecluster.
•Theschedulerqueuesthetaskssubmittedtothewholeclusterandassignsthetaskstotheprocessingnodes
accordingtonodeavailability.
•Thedistributedschedulerforthecloudapplicationhasspecialcharacteristicsthatcansupportcloud
applications,suchasschedulingtheprogramswritteninMapReducestyle.
•Theruntimesupportsystemkeepsthecloudclusterworkingproperlywithhighefficiency.

8. INTER-CLOUD RESOURCE MANAGEMENT
Resource Provisioning and Platform Deployment
•CloudarchitectureputsmoreemphasisonthenumberofprocessorcoresorVMinstances.
1.ProvisioningofComputeResources(VMs)
•ProviderssupplycloudservicesbysigningSLAswithendusers.
•TheSLAsmustcommitsufficientresourcessuchasCPU,memory,andbandwidththattheusercanusefora
presetperiod.
•UnderprovisioningofresourceswillleadtobrokenSLAsandpenalties.
•Overprovisioningofresourceswillleadtoresourceunderutilization,andconsequently,adecreaseinrevenue
fortheprovider.
•Deployinganautonomoussystemtoefficientlyprovisionresourcestousersisachallengingproblem.
•Thedifficultycomesfromtheunpredictabilityofconsumerdemand,softwareandhardwarefailures,
heterogeneityofservices(UsermaytakeNaaS,QueueService,SaaS),powermanagement(heatdissipationfromserver),and
conflictsinsignedSLAsbetweenconsumersandserviceproviders.

9. INTER-CLOUD RESOURCE MANAGEMENT
Resource Provisioning and Platform Deployment
1.ProvisioningofComputeResources(VMs)(Cont…)
•EfficientVMprovisioningdependsonthecloudarchitectureandmanagementofcloudinfrastructures.
•Inavirtualizedclusterofservers,thisdemandsefficientinstallationofVMs,liveVMmigration,andfast
recoveryfromfailures.
•TodeployVMs,userstreatthemasphysicalhostswithcustomizedoperatingsystemsforspecificapplications.
•Forexample,Amazon’sEC2(IaaSservicefromAmazon)usesXenasthevirtualmachinemonitor(VMM).Thesame
VMMisusedinIBM’sBlueCloud.
•IntheEC2platform,somepredefinedVMtemplatesarealsoprovided.Userscanchoosedifferentkindsof
VMsfromthetemplates.
•IBM’sBlueClouddoesnotprovideanyVMtemplates.Ingeneral,anytypeofVMcanrunontopofXen.
•MicrosoftalsoappliesvirtualizationinitsAzurecloudplatform.Theprovidershouldofferresource-economic
services.

10. INTER-CLOUD RESOURCE MANAGEMENT
Resource Provisioning and Platform Deployment
2.ResourceProvisioningMethods
➢Demand-Drivenmethod
➢EventDrivenmethod
➢Popularity-Drivenmethod

11. INTER-CLOUD RESOURCE MANAGEMENT
Resource Provisioning and Platform Deployment
Demand-Drivenmethod
•Thismethodaddsorremovescomputinginstancesbasedonthecurrentutilizationleveloftheallocatedresources.
•Thedemand-drivenmethodautomaticallyallocatestwoXeonprocessorsfortheuserapplication,whentheuserwas
usingoneXeonprocessormorethan60percentofthetimeforanextendedperiod.
•Ingeneral,whenaresourcehassurpassedathresholdforacertainamountoftime,theschemeincreasesthatresource
basedondemand.Whenaresourceisbelowathresholdforacertainamountoftime,thatresourcecouldbedecreased
accordingly.(DefinesarangeforCPUutilizationsayforeg:30%to70%.ifCPUutilizationbelow30%decreasestheCPUcapacity.IfCPUutilizationabove70%increasestheCPUcapacity)
•Amazonimplementssuchanauto-scalefeatureinitsEC2platform.
•Thismethodiseasytoimplement.
•Disadvantage:Theschemedoesnotworkoutrightiftheworkloadchangesabruptly.

12. INTER-CLOUD RESOURCE MANAGEMENT
Resource Provisioning and Platform Deployment
EventDrivenmethod
•Thisschemeaddsorremovesmachineinstancesbasedonaspecifictimeevent.
•TheschemeworksbetterforseasonalorpredictedeventssuchasChristmastimeintheWestandtheLunar
NewYearintheEast.
•Duringtheseevents,thenumberofusersgrowsbeforetheeventperiodandthendecreasesduringtheevent
period.
•Thisschemeanticipatespeaktrafficbeforeithappens.
•ThemethodresultsinaminimallossofQoS,iftheeventispredictedcorrectly.Otherwise,wastedresources
areevengreaterduetoeventsthatdonotfollowafixedpattern.

13. INTER-CLOUD RESOURCE MANAGEMENT
Resource Provisioning and Platform Deployment
Popularity-Drivenmethod
•Inthismethod,theInternetsearchesforpopularityofcertainapplicationsandcreatestheinstancesby
popularitydemand.(Currentlypopularapplications→Facebook,Instagram,Twitter)
•Theschemeanticipatesincreasedtrafficwithpopularity.
•Again,theschemehasaminimallossofQoS,ifthepredictedpopularityiscorrect.Resourcesmaybewastedif
trafficdoesnotoccurasexpected.

14. INTER-CLOUD RESOURCE MANAGEMENT
•(Gridisadistributedhighperformancecomputingparadigmthatoffersvarioustypesof
resources(likecomputing,storage,communication)toresource-intensiveusertasks.)Grid→a
sitewhichprovidessomesetofresourcesfortheuserapplications)
•ThecloudusesVMsasbuildingblockstocreatean
executionenvironmentacrossmultipleresourcesites.
•TheInterGrid-managedinfrastructurewasdeveloped
byaMelbourneUniversitygroup.
•TheInterGridisaJava-implementedsoftwaresystem
thatletsuserscreateexecutioncloudenvironmentson
topofallparticipatinggridresources.(Framework/Software
designedbyMelbourneUniversitygroupthatrunsoneachgrid(organizationsite/cloudsite)
thatallowsuserstocreateVMsonthetopofallparticipatinggrid,whereeachgridmaintains
thesetofresources)
Resource Provisioning and Platform Deployment
Dynamic Resource Deployment

15. INTER-CLOUD RESOURCE MANAGEMENT
•Peeringarrangementsestablishedbetweengateways
enabletheallocationofresourcesfrommultiplegridsto
establishtheexecutionenvironment.
•InFigure,ascenarioisillustratedbywhichanintergrid
gateway(IGG)allocatesresourcesfromalocalclusterto
deployapplicationsinthreesteps:(1)requestingthe
VMs,(2)enactingtheleases(sanctioning),and(3)
deployingtheVMsasrequested.
•Underpeakdemand,thisIGGinteractswithanotherIGG
thatcanallocateresourcesfromacloudcomputing
provider.
Resource Provisioning and Platform Deployment
Dynamic Resource Deployment

16. INTER-CLOUD RESOURCE MANAGEMENT
•Agridhaspredefinedpeeringarrangementswithothergrids,whichtheIGG
manages.
•ThroughmultipleIGGs,thesystemcoordinatestheuseofInterGridresources.
•AnIGGisawareofthepeeringtermswithothergrids,selectssuitablegridsthat
canprovidetherequiredresources,andrepliestorequestsfromotherIGGs.
•AnIGGcanalsoallocateresourcesfromacloudprovider.
•Thecloudsystemcreatesavirtualenvironmenttohelpusersdeploytheir
applications.Theseapplicationsusethedistributedgridresources.
•TheInterGridallocatesandprovidesadistributedvirtualenvironment(DVE).This
isavirtualclusterofVMsthatrunsisolatedfromothervirtualclusters.
•AcomponentcalledtheDVEmanagerperformsresourceallocationand
managementonbehalfofspecificuserapplications.
Resource Provisioning and Platform Deployment
Dynamic Resource Deployment

17. INTER-CLOUD RESOURCE MANAGEMENT
Resource Provisioning and Platform Deployment
Provisioning of Storage Resources

18. INTER-CLOUD RESOURCE MANAGEMENT
Virtual Machine Creation and Management
•FigureshowstheinteractionsamongVMmanagersforVMcreationandmanagement.Themanagersprovidea
publicAPIforuserstosubmitandcontroltheVMs.

19. INTER-CLOUD RESOURCE MANAGEMENT
Virtual Machine Creation and Management
IndependentServiceManagement
•Independentservicerequestfacilitiestoexecutemanyunrelatedtasks.
•Commonly,theAPIsprovidedaresomewebservicesthatthedevelopercanuseconveniently.
•InAmazoncloudcomputinginfrastructure,SQSisconstructedforprovidingareliablecommunicationservice
betweendifferentproviders.Eventheendpointdoesnotrunwhileanotherentityhaspostedamessagein
SQS.
•Byusingindependentserviceproviders,thecloudapplicationscanrundifferentservicesatthesametime.
(providingdata,computeorstorageservices).

20. INTER-CLOUD RESOURCE MANAGEMENT
Virtual Machine Creation and Management
RunningThird-PartyApplications
•Cloudplatformshavetosupportforbuildingapplicationsbyprovidingapplicationsthatareconstructedbythird-party
applicationprovidersorprogrammers.
•AscurrentwebapplicationsareoftenprovidedbyusingWeb2.0forms(interactiveapplicationswithAjax),the
programminginterfacesaredifferentfromthetraditionalprogramminginterfacessuchasfunctionsinruntimelibraries.
•TheAPIsareoftenintheformofservices.
•Webserviceapplicationenginesareoftenusedbyprogrammersforbuildingapplications.
•Asexamples,GAEandMicrosoftAzureapplytheirowncloudAPIstogetspecialcloudservices.
•TheWebSphereapplicationengineisdeployedbyIBMforBlueCloud.Itcanbeusedtodevelopanykindofweb
applicationwritteninJava.

21. INTER-CLOUD RESOURCE MANAGEMENT
Virtual Machine Creation and Management
VirtualMachineManager
•TheVMmanageristhelinkbetweenthegatewayandresources.
•Thegatewaydoesn’tsharephysicalresourcesdirectly,butreliesonvirtualizationtechnology.Hence,the
actualresourcesitusesareVMs.(VIErunsateachcloud,VMMconnectswithdifferentVIE.
•ThemanagermanageVMsdeployedonasetofphysicalresources.
•TheVMmanagerimplementationisgenericsothatitcanconnectwithdifferentVIEs.Typically,VIEscancreate
andstopVMsonaphysicalcluster.(IfOpenNebulaplatformrunningatonecloudtocreateVMs,thenVIEcancommunicatewithVIEofother
cloudrunningAmazonEC2platformtocreateVMs)
•TheMelbournegrouphasdevelopedmanagersforOpenNebula,AmazonEC2,andFrenchGrid’5000.
•TodeployaVM,themanagerneedstouseitstemplate

22. INTER-CLOUD RESOURCE MANAGEMENT
Virtual Machine Creation and Management
VirtualMachineTemplates
•AVMtemplateisanalogoustoacomputer’sconfigurationandcontainsadescriptionforaVMwiththefollowing
staticinformation:
➢ThenumberofcoresorprocessorstobeassignedtotheVM
➢TheamountofmemorytheVMrequires
➢ThekernelusedtoboottheVM’soperatingsystem
➢ThediskimagecontainingtheVM’sfilesystem(Files)
➢ThepriceperhourofusingaVM

23. INTER-CLOUD RESOURCE MANAGEMENT
Virtual Machine Creation and Management
VirtualMachineTemplates
•ThegatewayadministratorprovidestheVMtemplateinformationwhentheinfrastructureissetup.The
administratorcanupdate,add,anddeletetemplatesatanytime.
•Inaddition,eachgatewayintheInterGridnetworkmustagreeonthetemplatestoprovidethesameconfiguration
oneachsite.
•TodeployaninstanceofagivenVM,theVMMgeneratesadescriptorfromthetemplate.
•ThisdescriptorcontainsthesamefieldsasthetemplateandadditionalinformationrelatedtoaspecificVM
instance.
•Typicallytheadditionalinformationincludes:
➢ThediskimagethatcontainstheVM’sfilesystem
➢TheaddressofthephysicalmachinehostingtheVM
➢TheVM’snetworkconfiguration

24. INTER-CLOUD RESOURCE MANAGEMENT
Virtual Machine Creation and Management
Distributed VM Management

25. INTER-CLOUD RESOURCE MANAGEMENT
Global Exchange of Cloud Resources
•Inordertosupportalargenumberofconsumersfromaroundtheworld,cloudinfrastructureprovidershave
establisheddatacentersinmultiplegeographicallocationstoprovideredundancyandensurereliabilityin
caseofsitefailures.
•Forexample,AmazonhasdatacentersintheUnitedStates(e.g.,oneontheEastCoastandanotheronthe
WestCoast)andEurope.
•However,itisdifficultforcloudcustomerstodetermineinadvancethebestlocationforhostingtheirservices
astheymaynotknowtheoriginofconsumersoftheirservices.
•Also,SaaSprovidersmaynotbeabletomeettheQoSexpectationsoftheirserviceconsumersoriginatingfrom
multiplegeographicallocations.
•Thisnecessitatesbuildingmechanismsforseamlessfederationofdatacentersofacloudproviderorproviders
supportingdynamicscalingofapplicationsacrossmultipledomainsinordertomeetQoStargetsofcloud
customers.(CreatingofVMsatmultipledatacentersatmultipleplacesallovertheworldthatsatisfiescustomerQoS),

26. INTER-CLOUD RESOURCE MANAGEMENT
•Figureshowsthehigh-levelcomponentsoftheMelbourne
group’sproposedInterCloudarchitecture.
•Inaddition,nosinglecloudinfrastructureproviderwillbeableto
establishitsdatacentersatallpossiblelocationsthroughoutthe
world.
•Asaresult,cloudproviderswillhavedifficultyinmeetingQoS
expectationsforalltheirconsumers.
•Hence,theywouldliketomakeuseofservicesofmultiplecloud
infrastructureserviceproviderswhocanprovidebettersupport
fortheirspecificconsumerneeds.
•Thisnecessitatesfederationofcloudinfrastructureservice
providersforseamlessprovisioningofservicesacrossdifferent
cloudproviders.
Global Exchange of Cloud Resources

27. INTER-CLOUD RESOURCE MANAGEMENT
•Torealizethis,theUniversityofMelbournehasproposed
InterCloudarchitecturesupportingbrokeringandexchangeof
cloudresourcesforscalingapplicationsacrossmultipleclouds.
•Cloudproviderswillbeabletodynamicallyexpandorresize
theirprovisioningcapabilitybasedonsuddenspikesin
workloaddemandsbyleasingavailablecomputationaland
storagecapabilitiesfromothercloudserviceproviders;
operateaspartofamarket-drivenresourceleasing
federation.
•Theyconsistofclientbrokeringandcoordinatorservicesthat
supportutility-drivenfederationofclouds:application
scheduling,resourceallocation,andmigrationofworkloads.
Global Exchange of Cloud Resources

28. INTER-CLOUD RESOURCE MANAGEMENT
•TheCloudExchange(CEx)actsasamarketmakerforbringingtogether
serviceproducersandconsumers.Itaggregatestheinfrastructure
demandsfromapplicationbrokersandevaluatesthemagainstthe
availablesupplycurrentlypublishedbythecloudcoordinators.
•Itsupportstradingofcloudservicesbasedoncompetitiveeconomic
modelssuchascommoditymarketsandauctions.
•AnSLAspecifiesthedetailsoftheservicetobeprovidedintermsof
metricsagreeduponbyallparties,andincentivesandpenaltiesfor
meetingandviolatingtheexpectations,respectively.
•Theavailabilityofabankingsystemwithinthemarketensuresthat
financialtransactionspertainingtoSLAsbetweenparticipantsare
carriedoutinasecureanddependableenvironment.
Global Exchange of Cloud Resources