Curie

Description

Curie is a cluster featuring hybrid CPU/GPU nodes. It is equipped with the following hardware/software:

• 5 nodes with 2 x CPU AMD Opteron 6128 (16 cores), 16 GB of RAM, 1 x 250 GB HD SATA, 2 x GPU Tesla M2050, Infiniband, OS Scientific Linux 6.0.
• 2 nodes with 2 x CPU AMD Opteron 6128 (16 cores), 56 GB of RAM, 1 x 250 GB HD SATA, 2 x GPU Tesla M2075, Infiniband, OS CentOS 6.2.
• 3 nodes with 2 x CPU AMD Opteron 6238 (24 cores), 64 GB of RAM, 1 x 500 GB HD SATA, 3 x GPU Tesla K20Xm, Infiniband, OS CentOS 6.5.

for a total of 11 nodes, 184 cores, 23 GPU's.

Access

• Linux and Mac OS users can login within the DiSC internet connection using a terminal with the command:

ssh account@192.168.9.51

or from outside the Department using

ssh -p 50005 account@147.162.63.10

where "account" is the user's account.

• Windows users can use an interface program to the ssh service, e.g. Putty.

Queues

The queue manager is PBS and the following queues are available:

• curie: max nodes 4, max walltime 2 weeks (336 hours)
• curie2: max nodes 2, max walltime 2 weeks (336 hours)
• curie-test: max nodes 1, max walltime 10 minutes

Example PBS file

A typical PBS input file script will be as follow (see the Support page for more help):

#!/bin/sh --login
#PBS -N jobname
#PBS -A account
#PBS -q queue
#PBS -l nodes=n
#PBS -w time in hh:mm:ss format

commands to execute

where the parts in italic should be changed as appropriate.