above the clouds

Download Above the Clouds

Post on 25-Dec-2015




4 download

Embed Size (px)






  • OutlineWhat is it?Why now?Cloud killer appsEconomics for usersEconomics for providersChallenges and opportunitiesImplications


  • What is Cloud Computing?Old idea: Software as a Service (SaaS)Def: delivering applications over the InternetRecently: [Hardware, Infrastrucuture, Platform] as a servicePoorly defined so we avoid all X as a serviceUtility Computing: pay-as-you-go computingIllusion of infinite resourcesNo up-front costFine-grained billing (e.g. hourly) *

  • Why Now?Experience with very large datacentersUnprecedented economies of scaleOther factorsPervasive broadband InternetFast x86 virtualizationPay-as-you-go billing modelStandard software stack*

  • Spectrum of CloudsInstruction Set VM (Amazon EC2, 3Tera)Bytecode VM (Microsoft Azure)Framework VMGoogle AppEngine, Force.com*EC2AzureAppEngineForce.comLower-level,Less managementHigher-level,More management

  • Cloud Killer AppsMobile and web applicationsExtensions of desktop softwareMatlab, MathematicaBatch processing / MapReduceOracle at Harvard, Hadoop at NY Times


  • Economics of Cloud Users*Pay by use instead of provisioning for peakStatic data centerData center in the cloud

  • Unused resourcesEconomics of Cloud Users*Risk of over-provisioning: underutilizationStatic data center

  • Economics of Cloud Users*Heavy penalty for under-provisioningLost revenueLost users

  • Economics of Cloud Providers5-7x economies of scale [Hamilton 2008]

    Extra benefitsAmazon: utilize off-peak capacityMicrosoft: sell .NET toolsGoogle: reuse existing infrastructure*

    ResourceCost inMedium DCCost inVery Large DCRatioNetwork$95 / Mbps / month$13 / Mbps / month7.1xStorage$2.20 / GB / month$0.40 / GB / month5.7xAdministration140 servers/admin>1000 servers/admin7.1x

  • Adoption Challenges*

    ChallengeOpportunityAvailabilityMultiple providers & DCsData lock-inStandardizationData Condentiality and AuditabilityEncryption, VLANs, Firewalls; Geographical Data Storage

  • Growth Challenges*

    ChallengeOpportunityData transfer bottlenecksFedEx-ing disks, Data Backup/ArchivalPerformance unpredictabilityImproved VM support, flash memory, scheduling VMsScalable storageInvent scalable storeBugs in large distributed systemsInvent Debugger that relies on Distributed VMsScaling quicklyInvent Auto-Scaler that relies on ML; Snapshots

  • Policy and Business Challenges*

    ChallengeOpportunityReputation Fate SharingOffer reputation-guarding services like those for emailSoftware LicensingPay-for-use licenses; Bulk use sales

  • Short Term ImplicationsStartups and prototypingOne-off tasksWashington post, NY TimesCost associativity for scientific applicationsResearch at scale


  • Long Term ImplicationsApplication software:Cloud & client parts, disconnection toleranceInfrastructure software:Resource accounting, VM awarenessHardware systems:Containers, energy proportionality*



View more >