09 Cloud Computing

Embed Size (px)

Citation preview

  • 8/12/2019 09 Cloud Computing

    1/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    1Master of Information SystemManagement

    95-843 Service Oriented

    Architecture

    Cloud Computing and SOA

    Notes adapted from A View of Cloud Computing, Armbrust,and others -UC Berkeley Reliable, Adaptive, DistributedSystems Laboratory

  • 8/12/2019 09 Cloud Computing

    2/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Cloud Computing

    2Master of Information SystemManagement

    From A View of Cloud Computing, Armbrust, andothers -UC Berkeley Reliable, Adaptive, DistributedSystems Laboratory

  • 8/12/2019 09 Cloud Computing

    3/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Suppose you have aninnovative idea

    You need a large capital outlay inhardware.

    You need talented humans to operate

    and maintain the system. There is an over provisioning risk the

    new system may not be as popular asyou hoped.

    There is an under provisioning risk -missing and losing potential users.

    Cloud computing allows you to start

    small and grow as needed. 3Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    4/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Over or Under-Provisioning

    4Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    5/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Over or Under-Provisioning

    5Master of Information SystemManagement

    Shaded area is unused capability.Shaded area representsrequests not served.

    Less andless

    demand.

  • 8/12/2019 09 Cloud Computing

    6/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Master of Information SystemManagement

    Real world estimates

    Average server utilization is 5% to20%.

    Peak workload exceeds the average byfactors of 2 to 10.

    Users provision for the peak.

    Peak loads may occur based on the

    time of day or based on other factors(e.g. photo sharing after the holidays,drop/add within two weeks of start ofterm, etc.)

    6Master of Information System

  • 8/12/2019 09 Cloud Computing

    7/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Or, suppose you have as

    large batch-oriented task? How do we benefit from using a

    cloud?

    1000 servers for one hour costsnot more than 1 server for 1000hours.

    This degree of elasticity may beunprecedented in the history of IT.

    7Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    8/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Three new aspects

    The illusion of infinite computingresources on demand (no farahead provisioning concerns)

    The elimination of up frontcommitment by cloud users (startsmall and grow)

    Pay for resources on a short termbasis as needed (rewardconservation)

    8Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    9/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Key Enablers

    Construction and operation ofextremely large-scale commodity-computer datacenters at low cost

    locations

    Statistical multiplexing to increaseutilization (to each according to his

    needs) varies over time thisdiffers from traditional hosting

    Virtualization of computation,

    storage, and communication 9Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    10/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Three Examples

    AppEngine(Google) Build scalable webapplications fast. Not for general purposecomputing.

    Azure(Microsoft) Use .NET and .NETlibraries as needed. General purposecomputing on a Microsoft platform.

    EC2(Amazon) Elastic Compute Cloud

    (Choose OS and the entire software stack.General purpose computing

    10Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    11/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Three Examples

    AppEngine(Google) Least flexible

    Application domain-specific platform

    Automatic scaling and high availability

    Proprietary megastore for data storage Azure (Microsoft) Moderately Flexible

    Language independent softwaredevelopment platform

    EC2(Amazon) Highly Flexible

    Hardware virtual machine

    You build from the kernel upward

    11Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    12/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    1.Business continuity and service

    availability

    Will the cloud provider remain inbusiness over the long haul?

    Can we prevent a single point of

    failure by using more than oneprovider?

    12Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    13/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    2. Data Lock-in

    Cloud storage is essentially

    proprietary.

    SaaS developers cannot easily extract

    their data and place it on multiple

    clouds.

    One solution would be to

    standardize data APIs.

    13Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    14/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    3. Data Confidentiality/Auditability

    Security is one of the most

    often-cited objections to cloudcomputing.

    Think HIPAA regulations and

    Sarbanes-OxleyThis is an old and difficult

    problem but, in the cloud, more

    parties are involved. 14Master of Information SystemManagement

    The authorsbelieve this isno moredifficult than

    in traditionaldatacenters.

  • 8/12/2019 09 Cloud Computing

    15/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    4. Data transfer bottlenecks

    15Master of Information SystemManagement

    At $100 to $150 per terabyte transferred, costs may quicklyadd up. Shipping disks may be cheaper.

    In 2010, the Library of Congress claims to hold 235 terabytes.Wikipedia holds about 5.87 terabytes.Watson, of Jeopardyfame, has 16 terabytes of RAM.

    One terabyte is about 1012bytes.

  • 8/12/2019 09 Cloud Computing

    16/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    5. Performance unpredictability

    16Master of Information SystemManagement

    Multiple virtual machines can share CPUs and main memoryvery well in cloud computing. Network and disk I/Osharing is less predictable.

    Virtual machines are not new. IBM developed much of this inthe 1980s.

    Flash memory preserves information when powered offlike mechanical hard disks. It is much faster to accessand its use may reduce I/O interference.

  • 8/12/2019 09 Cloud Computing

    17/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    6. Scalable storage

    17Master of Information SystemManagement

    An open research question is to create a storage systemthat would not only meet existing programmer expectationsin regard to durability, high availability, and the ability tomanage and query data but combine them with the cloudadvantages of scaling arbitrarily up and down on demand.

    AWS uses Amazon Elastic Block Store (EBS) and SimpleDBAzure uses SQL Data Services and Azure Storage ServiceAppEngine uses Megastore/BigTableMany open source NO(Not Only)SQL Projects

  • 8/12/2019 09 Cloud Computing

    18/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    7. Bugs in large scale distributed

    systems

    News on April 25, 2011:Amazon said it has completed its recovery efforts anda tiny percentage of data lost won't be fully restored.

    "We're in the process of contacting these customers,"

    Amazon said Monday afternoon on its Web site.

    Amazon said it is "digging deeply" into the root causes

    behind last week's shutdown and will provide a

    "detailed post mortem" in the future.

    18Master of Information SystemManagement

  • 8/12/2019 09 Cloud Computing

    19/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    8. Scaling quickly

    19Master of Information SystemManagement

    Pay as you go applies to storage and network bandwidth.Simply count byes used.

    Google App Engine automatically scales based on loadincreases and decreases.

    AWS charges by the hour for the number of instances youoccupy even if your machine is idle.

    The challenge is to scale fast (up and down) to save moneywithout violating SLAs.

  • 8/12/2019 09 Cloud Computing

    20/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    9. Reputation fate sharing

    20Master of Information SystemManagement

    One bad apple spoils the bunch.

    EC2 IP addresses have beenblacklisted by spam prevention services blocking some goodguys from sending email.

    The FBI raided a Dallas data center because a company whoseservices were hosted there was under investigation. Manyinnocent bystanderswere down for days and some wentout of business.

  • 8/12/2019 09 Cloud Computing

    21/22

    95-843 SOAAdapted fromA View of Cloud ComputingUC Berkeley

    Top 10 Obstacles To CloudComputing

    10. Software licensing

    21Master of Information SystemManagement

    Current practice is to restrict the computer on which thesoftware can run. The software is purchased along with amaintenance fee. This is not the pay-as-you go model.

    Microsoft and IBM are now offering pay-as-you go pricing.

    In 2011, IBM charges $6.39 per hour for WebSphere withLotus Web Content Management running on EC2.

  • 8/12/2019 09 Cloud Computing

    22/22

    95-843 SOAAdapted fromA View of Cloud Computing

    SOA and Cloud Computing

    22Master of Information SystemM t

    While you can certainly leverage a cloud without practicing SOA,and you can leverage SOA without leveraging cloud computing,the real value of cloud computing is the ability to use services,data, and processes that can exist outside of the firewall

    From Cloud Computing and SOA Convergence in your Enterpriseby David S. Linthicum