Please enable JavaScript.

Coggle requires JavaScript to display documents.

Week 5 Distributed (Grids (Features (Entertainment (Gaming, File…

- - - - Expensive instruments
      - High speed links
      - Sensors
      - Multimedia
      - Data stores
    - - Some apps only available on certain platforms
      - Has QoS capability
      - Production ready infrastructure
      - Runs specialist applications
      - Uses interoperable protocols
  - - - Remote access to expensive scientific instruments
    - - Pay per use grids
    - - Gaming
      - File distribution
  - - - Authentication of users/services
      - Authorisation of comms
      - Manage creds and group membership
    - - Secure connection over internet
      - Virtualisation
      - Scalability
  - - - Does not hide architecture of components
      - Better utilisation of underlying components
      - Apps require infrastructure level awareness (costly)
      - Better message passsing libraries
    - - Has JVM
      - Cannot guarantee use of all underlying infrastructure
      - Easy to develop on
- - - - Utility computing service
      - Self-service
        
        Virtual machines
        
        Storage
        
        Network configs
      - Allow employees of company to set up their own cloud
    - - Made up of several different projects
      - Scalable operating system for clouds
    - - Open source
      - Create AWS compatible clouds
- - - - Additional layer of abstraction over native fs
    - - Has several Datanodes
        
        Stores portion of the filesystem
      - Has a central Namenode
        
        Stores two copies of each block for redundancy on different datanodes
        
        RPC server for group of Datanodes
        
        Main point of overload
      - Maintain copy of FsImage and EditLog for block mapping and journaling
- - - - Clusters running MPI
      - Grids running MPI or other parallel apps
    - - Clusters for web serving
      - Grids with mixed apps
      - Clouds with mixed work
    - - Specialized apps
    - - PVM (Parallel Virtual Machine) software
      - MPI (Message Passing Interface) software
      - Parametric Computing Toolsets (Nimrod/Enfuzion)
      - NFS (Network File System)
      - AFS (Andrew File System)
        
        Greater fault tollerance
        
        Better resource utilization
      - Hadoop
      - Spark
    - - MPP
        
        Limited to embarrassingly parallel methods
        
        Image processing
        
        Cryptography
        
        Parametric computing
        
        Expensive to administer
        
        Costly due to custom hw
        
        Highly application specific
      - SMP
        
        Difficult to scale due to bus limitations
        
        Low parallelism due to bus size
        
        Provides transparent access to parallelism
        
        Costly because of custom HW
        
        Widely used in period server hosts