Galaxy product concepts in review

   * Linear performance over a large # of CPUs (6+)

   * Shared memory for application usage across instances

   * Multiple computing models (Shared Nothing/Shared Everything)

   * Independent instances of operating systems

   * Dynamic reconfiguration on the fly (no reboot)

   * Hot swapping

   * Upgrades

   * Scalable I/O

   * CPU Migration (Drag & Drop)

   * Can implement multiple simultaneous computing models

 Previous                  Back to first       View graphic
 slide        Next slide   slide               version

Notes:
----------------------------------------------------------------------------

   * Software scalability for large multi-CPU systems
        o 1 to 256+ CPUs (32 nodes, 8 CPUs per node)
        o 256MB to 1+ TB physical memory
        o 1 to 500+ I/O ports
        o 10 to 1000’s of disks

   * 300,000+ TPM high-end design center
        o 1,000,000 TPM goal

   * Dynamic resource allocation (CPU, Memory)

   * Cluster Global Sections (Multi-Gigabyte)

   * Hot swapping and upgrades of CPUs, Memory, and I/O

     components without shutting down entire system