High energy & nuclear physics


Some convergence between P2P and grid computing



Yüklə 446 b.
səhifə9/9
tarix01.11.2017
ölçüsü446 b.
#24870
1   2   3   4   5   6   7   8   9

  • Some convergence between P2P and grid computing

  • The industrialization time



  • 2008 conclusion is still valid

    • 2008 conclusion is still valid

    • … except that cloud computing has emerged!

    • Will cloud computing kill grid computing? No!

    • Is cloud computing just an avatar of grid computing? Yes and No

    • Let’s have a closer look at cloud computing ;-) !





    In the early stage of grid computing, its promoters (esp. I. Foster) claimed that it would dramatically change the society and its development, both in its economic and social dimensions

    • In the early stage of grid computing, its promoters (esp. I. Foster) claimed that it would dramatically change the society and its development, both in its economic and social dimensions

      • Analogy with previous networks which shaped the development of the US: railways, telephone, electricity, roads, bank networks
      • Argue for similar characteristics and issues: intrinsic complexity, use of/need for standards, distribution, integration (large/small)
    • Most Governments were convinced and invested billions of $/£/€

    • However, important differences were under-estimated

      • P2P/partner vs Customer-provider relationship
      • Large number vs Small number of actors/providers
      • Large number vs Small number of applications
      • Weak vs Strong supervision/control


    Consequence: this forecast failed… and governments were angry

    • Consequence: this forecast failed… and governments were angry

    • Not a complete failure: grids are definitely useful and actually used!

    • But they did not re-shape the economy and the society

    • However they did push extremely strong ideas

      • Aggregation of distributed resources
      • Heavy use of clusters
      • Brokering of resources – Large scale monitoring of distributed resources
      • World-Wide scale
      • Access through the Internet
    • Merge these ideas with ASP (Application Service Providers)/ outsourcing and Service Oriented Architecture (SOA), and we will get the picture of cloud computing



    Cloud computing have old roots

    • Cloud computing have old roots

      • John Mac Carthy (1962): “Computation may someday be organized as a public utility”
      • Douglas Parkhill (1966): “The Challenge of the Computer Utility”
      • J.C.R. Licklider (1969): “Intergalactic computer network“
      • until the mid-eighties, model = 1 mainframe + n terminals – CPU time sold as a commercial good
    • Key issues (related to some form of QoS) needed to be solved

      • The network speed, esp. the bandwidth
      • The storage capacity
    • They issues now are more or less solved



    First considered-as-a-cloud infrastructure: Salesforces.com (1999) (delivering of business applications from a website)

    • First considered-as-a-cloud infrastructure: Salesforces.com (1999) (delivering of business applications from a website)

    • Amazon Web Services: 2002

    • Amazon EC2: 2006

    • IBM Blue Cloud: 2007

    • Google Apps: 2009



    A supposed-to-be user-centered vision

    • A supposed-to-be user-centered vision

      • managing a computer is exhausting
      • the user does not care about the system components: the user just want his problem to be solved
      • eliminate the burden of the software/hardware management
      • accelerate application launches – facilitates business evolutions
      • allow the user benefit from economies of scale… and fire computer engineers!
    • A business/market vision

      • marketing existing under-used resources
      • a small set of computing power providers
      • a global market
      • an integrated « hyper-market »: computing, entertainment, learning?
      • for the best of a few big (too-big-to-fail) companies


    "A Cloud is a type of parallel and distributed system consisting of a collection of interconnected and virtualized computers that are dynamically provisioned and presented as one or more unified computing resources based on service-level agreements established through negotiation between the service provider and consumers” (Buyya et al.)

    • "A Cloud is a type of parallel and distributed system consisting of a collection of interconnected and virtualized computers that are dynamically provisioned and presented as one or more unified computing resources based on service-level agreements established through negotiation between the service provider and consumers” (Buyya et al.)

    • “A large-scale distributed computing paradigm that is driven by economies of scale, in which a pool of abstracted, virtualized, dynamically-scalable, managed computing power, storage, platforms, and services are delivered on demand to external customers over the Internet” (Foster et al.)



    Excerpts from the definitions

    • Excerpts from the definitions

      • large-scale distributed computing paradigm
      • driven by economies of scale
      • collection of interconnected and virtualized computers
      • dynamically provisioned - delivered on demand
      • presented as one or more unified computing resources
      • service
      • provider and consumers - service-level agreements
      • over the Internet
    • Other important characteristics

      • reliability
      • performance
      • security


    http://www.ibm.com/cloud-computing/us/en/

    • http://www.ibm.com/cloud-computing/us/en/

    • (2008) « Values to customers include:

      • Reducing IT management complexity and skill requirements
      • Sharing resources among multiple applications
      • Accelerating application launches
      • Supporting both existing and emerging, data-intensive workloads
    • »

    • (2013) « Harness the power of cloud: the power of cloud enables you to supercharge innovation, gain agility, capitalize on analytics, and get personal:

      • Deliver real-time personalized services
      • Gain intelligence by connecting networks
      • Become nimble and agile – evolve fast
      • Capture new markets and opportunities
    • »



    From the customer point of view

    • From the customer point of view

      • Scalability
      • Reliability
      • Flexibility
      • Cost
      • Security and Privacy
      • Performance
      • Ubiquitous and fast access
      • Quality of Service
      • Service Level Agreement
      • Pricing system
      • Simple to use
    • From the internal point of view

      • Virtualization
      • « Grid » management


    Public clouds (or external clouds)

    • Public clouds (or external clouds)

      • resources are dynamically provisioned on a fine-grained, self-service basis over the Internet, via Web applications or Web services.
      • the cloud is hosted, operated, and managed by a third-party vendor from one or more data/computing centers
      • organizations/customers lease shared resources from public clouds, effectively becoming infrastructure tenants rather than owners
      • computing becomes a public utility
    • Private clouds (or internal clouds) emulate cloud computing on private networks

    • Hybrid clouds merge multiple internal and/or external clouds



    Infrastructure as a Service (IaaS)

    • Infrastructure as a Service (IaaS)

      • provides computing environments/infrastructure i.e., hardware, OS…
      • the provisioned infrastructure can dynamically scale up and down depending on the actual needs
      • Ex: Amazon EC2 and S3
    • Platform as a Service (PaaS)

      • high-level integrated environment to build and deploy applications: execution runtime, DBMS, web server…
      • restrictions on the type of applications
      • but scalable platform
      • Ex: Google’s App Engine for deploying Web applications
    • Software as a Service (SaaS)

      • delivers software to consumers through the Internet
      • Ex: Salesforce: online CRM (Customer Relationship Management) Services ; Live Mesh from Microsoft: files and folders synchronization; email services




    Storage as a Service 

    • Storage as a Service 

    • Data(base) as a Service (DaaS)

    • Process as a Service - Business Process as a Service (BPaaS)

    • Network as a Service (NaaS)

    • Integration as a Service 

    • Security as a Service - Identity as a Service

    • Testing as a Service

    • EAAS: Everything as a Sevrvice



    Core (Physical) Infrastructure

    • Core (Physical) Infrastructure

      • Data centers
      • Clusters of “computing elements”
      • Network
    • Virtualization Management/Orchestration

    • Backup and Recovery

    • Monitoring, Load Balancing, and Brokering (Provisioning)

      • Performance and event monitoring
      • Brokering and supervision of resources


    Security Management

    • Security Management

      • Identity management
      • Access control
      • Security infrastructure (firewall, IPDS, anti-virus, encryption servers…)
    • Accounting and Service Level Management

    • Information Management

      • Service catalog
      • Resource catalog


    Virtualization and Hypervisors

    • Virtualization and Hypervisors

      • hide the physical hardware – simulate the hardware: virtual machine
      • allows the execution of multiple OS on the same hardware
      • hypervisor: manages virtual machine monitors and execute base functions (e.g. scheduling)
      • full virtualization (e.g., VMWare): include binary translation of non-virtualizable (e.g., privileged instructions in x86) instructions
      • paravirtualization (e.g., Xen): modify the OS + hyper-calls to the hypervisor to replace non-virtualizable instructions
      • hardware-assisted virtualization: WM monitor executed in a specific “root” mode + virtualization tasks in hardware + control of the I/O by the hypervisor




    Compute Cloud EC2 (Elastic Cloud Computing)

    • Compute Cloud EC2 (Elastic Cloud Computing)

      • pricing: http://aws.amazon.com/fr/ec2/pricing//185-8840065-1558013/
      • « private » virtualized servers (« instances ») of different types
      • example: High-CPU Extra Large Instance
        • 7 GB of memory
        • 20 EC2 Compute Units
        • 1690 GB Storage
      • pricing on a per hour basis for each instance type
        • from $0.065/hour for the small standard "On-Demand" virtual machine running Linux to 8x more for the largest one running Windows (Oct. 2013-EU prices)
        • data transfer charge ranges from $0 to $0.12 per GB-month, depending on the volume
      • (note : prices in 2011: $0,085/hour upto 29x - transfer price: $0,08 to 0,15!!!)


    • Data Cloud S3 (Simple Storage Service)

      • pricing: http://aws.amazon.com/fr/s3/pricing/
      • from $0.055 to $0.095 per GB-month (standard storage)
      • + bandwidth usage (from $0 to 0.12 per GB – EU price)
      • + requests (from $0.005 to $0.055 per 1000 requests)
      • (note: prices in 2011: storage: 0,055 to 0,14 – transfer: $0.05 to 0.12 – requests: $0.001 to $0.01 )
      • data transfer is charged by TB / month data transfer, depending on the source and target of such transfer
      • 2000 billions stored objects (Apr. 2013)
      • (Oct. 2011: 556 billion; March 2010: 102 billion objects




    Computing

    • Computing

      • connection to the « Compute » service > 99,95%
      • Web and Worker roles > 99,9%
      • note : service interruption  interruption > 5 mn
    • Storage

      • « failed » transactions (error rate) < 0,1%
      • credit = 10%
        • if 0,1% <= error rate < 1%
      • Credit = 25%
        • if error rate >= 1%






    Database service

    • Database service

      • https://www.windowsazure.com/fr-fr/pricing/details/sql-database
      • costs is highly dependent from the size of the DB
      • ~1€ per GB/month
      • not so cheap!
    • Comparison with Oracle pricing (Oct. 2013)

      • Standard Edition
        • 350€ per named user + 77€ maintenance
        • or 17,500 € (no limit-processor license)+ 3,850€ maintenance
      • Enterprise Edition
        • 950€ per named user + 209€ maintenance
        • or 47,500 € (no limit-processor license) + 10,450€ maintenance


    For one computer

    • For one computer

      • From $4 to $5 per month per computer!
    • Unlimited data size

    • Possible encryption

    • From 20 TB 5 years ago to 40 PB in 2013



    Previous previsions

    • Previous previsions

      • Gartner (2010):
        • $150.1 billion by 2013
        • identifies the Cloud as one of the four trends that will change IT and the economy in the next 10 years (http://www.gartner.com/it/page.jsp?id=1470115)
      • IDC (2010):
        • the market for private enterprise Cloud servers will grow from an $8.4 billion opportunity in 2010, to a $12.6 billion market in 2014
        • SaaS revenue will grow five times more than traditional software
        • by 2014, about 34% of all new business software purchases will be consumed via SaaS
    • 2012 key figures (Gartner, Feb. 2013)

      • revenue: ~$130 billion (public clouds) (+18,5%)
      • IAAS : +43% ($9 billions) (fastest-growing segment)
      • cloud advertising: 48%!!! ($62 billon!) (1st segment)
      • BPaaS: 28% - SaaS: 14,7%; PaaS: only 1%
      • Prediction for 2016: $677 billon, incl. $310 billion for advertising (+17,7% of annual growth)
    • Cloud vendors are experiencing growth rates of 90% per annum (FSN, March 2013)





    Steve Ballmer (Microsoft's CEO) (March 2010)

    • Steve Ballmer (Microsoft's CEO) (March 2010)

      • about 75% (90% in a year) of Microsoft customers are doing entirely cloud based or entirely cloud inspired
    • Real adoption of cloud: yet the early days!

      • 75% according to RightScale (cloud service broker) – strong biais: RightScale customers!
      • 30% according to Forrester Research
      • 37% according to Virtustream/Neovise (cloud platform providers)


    56% of UK CIOs and Senior IT leaders see complexity of their own ICT systems as the biggest barrier to their adoption of the cloud (NTT report, May 2013)

    • 56% of UK CIOs and Senior IT leaders see complexity of their own ICT systems as the biggest barrier to their adoption of the cloud (NTT report, May 2013)

      • 53%: launching new services and applications more quickly (80% in the transport and logistics sectors
      • 60%: cloud providers do not  appreciate how complex legacy ICT systems are, and fear migration to the cloud could fail
      • 46%: cloud is a great enabler of ‘bring your own device’ and flexible working, through enabling remote access to data and applications
      • Challenge: making cloud infrastructure work seamlessly with legacy platforms and applications
      • 28%: legacy systems are too expensive (or valuable) to abandon 
      • 68% have had cloud-based systems in place for two years or less




    Integration of legacy systems

    • Integration of legacy systems

    • Limited customization

    • Complexity

    • Lack of standards

    • (almost) Impossible migration/reversion

    • Security, Privacy

    • Legal issues and regulations

    • Reliability

    • Performance

    • Financial Cost

    • Reorganization cost

    • Human cost





    Collaborative computing, virtualization, dynamic provisioning of resources have definitely demonstrated their value

    • Collaborative computing, virtualization, dynamic provisioning of resources have definitely demonstrated their value

      • from a high performance computing point of view
      • from a cost effectiveness point of view
      • from a business point of view
    • No one-size-fits-all solution → hybrid approaches

    • Still strong open issues

    • A domain prone to bubbles



    Yüklə 446 b.

    Dostları ilə paylaş:
    1   2   3   4   5   6   7   8   9




    Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur ©muhaz.org 2024
    rəhbərliyinə müraciət

    gir | qeydiyyatdan keç
        Ana səhifə


    yükləyin