Sgees018

From SGEES Wiki
Jump to: navigation, search

High Performance cluster sgees018.geo.vuw.ac.nz

  • 80 core Intel(R) Xeon(R) CPU E5-2698 v4 @ 2.20GHz
  • 2 GPU NVIDIA Corporation GK210GL [Tesla K80]
  • 256 GiB Memory
  • Fast network (2x1 Gbit bonding interface)
  • Virtually unlimited ITS storage

Operating system

Ubuntu 16.04 LTS

Access

From the university network just use full-qualified host name for access the server: ssh -Y username@sgees018.geo.vuw.ac.nz

Or run X2Go client for starting remote XFCE session on the server.


Multi-Core Usage

Matlab

Matlab has multiprocessor and multithread ability. To check if it's true type in Matlab command window:

  maxNumCompThreads

Further reading about Matlab multicore:

MPD

To run multithreaded job you need to run MPD daemon in prior. By default MPD knows nothing about hardware so you should help it by run as:

mpd --ncpus=64

Performance and processes monitoring

Start another shell at sgees018 (screen, tmux, another SSH session) and monitor load and performance with htop

Job Managing

See Job Managing