You are here: Home / Cluster

ECS Cluster Computer

Cluster Overview

  • ecs-cluster.serv.pacific.edu
  • Cray (Appro) CS300-AC cluster computer
  • 8 nodes housed in Cray (Appro) GreenBlade SR5110 Chassis
  • Compute Node configuration:
    • Cray (Appro) GB512X
    • Two Intel E5-2695 v2 processors
    • 48 cores/node
    • 64GB RAM/node
    • Diskless/stateless nodes (except for node1)
  • GPU Node configuration
    • Two Nvidia K20X GPUs
    • Intel "Grizzly Pass" R2308GZ4GS9 motherboard
    • CPU:  Two Intel Xeon E5-2620 v2 processors
    • RAM: 64GB DDR3-1866
  • Networking:
    • 40 Gigabit Ethernet (Mellanox)
    • Gigabit Ethernet
  • Storage:
    • OS/Application Storage:  120GB SSD on node1
    • Primary-Data:  4TB SSD, ZFS Array on FreeNAS server, 40Gbit Ethernet
    • Backup-Data:  8TB Disk, RAID-5 array in QNAP TS-470 PRO NAS, 1Gbit Ethernet
  • Software:
    • Linux, CentOS 7.3
    • xCAT Cluster Toolkit
    • SLURM resource manager

  

Architecture:

Cluster Architecture

Scheduling

  • "Compute" partition - 7 nodes, 48 cores per node
    • This is the default partition for long-running computation
  • Other partitions can be created for resource reservations if needed
  • Node1 is reserved for interactive tasks
 

Documentation

Tutorials:

Cluster Tools: