Zephyr and Phoenix Cluster Configurations

From Template
Jump to: navigation, search

The TRACC Zephyr cluster consists of

  • 88 compute nodes in a queue, each with:
    • 2 16-core AMD Interlagos 6273, 2.3 GHz CPUs
    • 1 TB of local scratch space
    • 32 GB of RAM
    • 1 InfiniBand QDR interconnect
  • In addition we also have a queue with two nodes that have 64GB on each node and a queue with two nodes that have 128GB on each node
  • 2 login nodes, each with
    • 2 16-core AMD Interlagos 6273, 2.3 GHz CPUs
    • 2 TB of RAID1 local scratch space
    • 32 GB of RAM
    • 1 InfiniBand QDR interconnect
  • A high performance Lustre-based file system, consisting of
    • 1 I/O nodes
    • 1 Dual Redundant storage controllers
    • 48 3TB hard drives/RAID6
    • 120 TB formatted user capacity
  • Two Administrative Nodes
  • One Applications node for user development
  • One Statistics gathering node
  • Five SuperMicro Gigabit Ethernet switches connected to the Argonne network with two 10 Gigabit Ethernet fiber links
  • A QLogic InfiniBand switch

All nodes are connected to one another with both Gigabit Ethernet and Infiniband.

All the nodes run CentOS Linux 6.x.

The TRACC Phoenix cluster consists of

  • 128 compute nodes, each with
    • 2 quad-core AMD 2378 Opteron CPUs
    • 200 GB of local scratch space
    • 8 GB of RAM
    • 1 InfiniBand DDR interconnect
  • 3 login nodes, each with
    • 2 dual-core AMD 2240 Opteron CPUs
    • 180 GB of local scratch space
    • 16 GB of RAM
    • 1 InfiniBand DDR interconnect
  • a high performance GPFS file system, consisting of
    • 4 I/O nodes
    • 1 DDN storage storage controller
    • 480 500 GB hard drives
    • 180 TB formatted capacity
  • a Foundry SX 800 Gigabit Ethernet switch
    • connected to the Argonne network with a 10 Gigabit Ethernet fiber link
  • a SilverStorm 9120 InfiniBand switch

All nodes are connected to one another with both Gigabit Ethernet and Infiniband.

All the nodes run Red Hat Enterprise Linux 4.5.