Home

The purpose of this wiki is to collect information for using cluster.

Overview

This information needs confirmation, and more information may be collected in the future

• 4 head node: sug-login1 ~ sug-login4, each has 4 cores, 8G memory
• 2 head node (more memory): sug-app1 ~ sug-app2, 4 cores, 16G memory, can be only accessed through login node (e.g. ssh sug-app1 within sug-login1)
• x compute node (not sure): each has 24 cores and ~200G memory

Queues

There're different queues on the cluster, but I think we're only allowed to run on the analysis queue? You may check the queue status using qstat -Q

Submit jobs

For most operations the qsub command is used to submit a script which will be run on the cluster. Some useful Torque parameters (you may include these options in the header of your script):

#!/bin/bash

#PBS -N job_name            ### specify job name
#PBS -l nodes=1:ppn=1       ### specify number of nodes and cores requested (default is 1?)
#PBS -l mem=2048mb          ### specify amount of memory (default is ?)
#PBS -q analysis            ### specify queue
#PBS -V                     ### export your current environment parameters to the job
#PBS -d directory           ### change the working directory
#PBS -e std.err             ### redirect stderr to this file
#PBS -o std.out             ### redirect stdout to this file
#PBS -A project_code        ### specify project code
#PBS -j oe                  ### Joins standard output and standard error
#PBS -t 1-10                ### define array jobs (in this case will run 10 jobs), use $PBS_ARRAYID to get access to the array index

If you want to check stdout and stderr in real time, do qsub -k oe when submitting the job

Storage