Introduction

Cluster

  • all nodes: Dual socket SandyBridge E5-2630L, with 64 GB @1333 RAM per node. BIOS: HT disabled, performance mode, Minimum C state: C6

  • all nodes: 10 Gb iWARP Ethernet card: used for data plane (MPI) and control plane (CERN public network, AFS)

  • lxbrf61c08 node has a directory /data which is NFS exported to all compute nodes. Mounted as /data on all compute nodes.

  • front-end node: lxbrf61c08

  • compute nodes: lxbrf61c01 lxbrf59c03 lxbrf61c04 lxbrf61c05 lxbrf61c06 lxbrf61c07 lxbrf57c02 lxbrf57c04 lxbrf57c05 lxbrf57c06 lxbrf57c07 lxbrf57c08

Measurement setup

  • MPI libary: mvapich2-1.8-1.slc6.x86_64
  • MPI benchmarks: osu-micro-benchmarks-4.0.1

Latency

  • OSU MVAPICH2 benchmark: osu_latency

  • matrix diagonal: intra-node MPI (shared memory)
  • non-diagonal: inter-node MPI (10 Gb iWARP)

  • techlab_iwarp_latency_measeurement_mvapich2.png:
    techlab_iwarp_latency_measeurement_mvapich2.png

  • techlab_iwarp_latency_measeurement_mvapich2_2.png:
    techlab_iwarp_latency_measeurement_mvapich2_2.png

Bandwidth (unidirectional)

  • OSU MVAPICH2 benchmark: osu_bw

  • matrix diagonal: intra-node MPI (shared memory)
  • non-diagonal: inter-node MPI (10 Gb iWARP)

  • techlab_iwarp_bw_measeurement_mvapich2.png:
    techlab_iwarp_bw_measeurement_mvapich2.png

  • techlab_iwarp_bw_measeurement_mvapich2_2.png:
    techlab_iwarp_bw_measeurement_mvapich2_2.png
Edit | Attach | Watch | Print version | History: r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r1 - 2016-05-26 - AritzBrosaIartza
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    HardwareLabs/HardwareLabsPublic All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback