Server

To access our server, make sure that you are connected to the UTD network:

If you’re on-campus, make sure that you are connected to the CometNet internet
If you’re working off-campus, make sure that you are connected to the UTD VPN by following the instructions here

Map Network Directory

To map the kenrod/ drive (formerly, the shared/ drive) to your local computer:

For Mac users, go to Finder > Go > Connect to Server, enter the following and press Connect:

smb://cvl-share.cvl.utdallas.edu/kenrod

For Windows users, follow the instructions here and enter the following for Folder field:

\\cvl-share.cvl.utdallas.edu\kenrod

Login with Command Line

For Max users, we can use the Terminal app, which is already included in Macs

For Windows users, install either Putty or MobaXterm

You can login by typing the following:

ssh -Y <user>@<hostname>

User is your UTD Net ID and hostname can be one of the following:

cvlkrcompute1.utdallas.edu or ponyo.utdallas.edu
cortex.cvl.utdallas.edu

Software

To see available software, run:

module avail

To load available software, run:

module load <software>

To unload software, run:

module unload <software>

Parallel Jobs

To run parallel jobs, we can use SLURM (preferred) or Sun Grid Engine (old).

Parallel Script Template

SLURM

To use SLURM, first access the server via cortex.cvl.utdallas.edu.

Basic SLURM commands:

sbatch <script>          # Submit a job
squeue                   # Show see pending jobs
scancel <jobid>          # cancel specific job
scancel -u <username>    # Cancel all jobs

Common SLURM settings:

# ------------------------------------------------------------------------------
# slurm settings
# ------------------------------------------------------------------------------
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=1        # can be increased if needed
#SBATCH --mem=4G                 # can be increased if needed
#SBATCH --partition=kenrod
#SBATCH --output=jid-%A_jname-%x.log
#SBATCH --mail-type=FAIL
#SBATCH --time=2-00:00:00        # day-hours:minutes:seconds format
#SBATCH --export=NONE            # ALL or comma-separated environment variables

Command Example:

sbatch /path/to/script.sh

Sun Grid Engine

To use Sun Grid Engine, run the following on cvlkrcompute1.utdallas.edu (or ponyo.utdallas.edu):

module load sge

Basic SGE commands:

qsub <script>    # Submit a job
qstat            # Show statuses of jobs
qdel <jobid>     # Cancel a job
qhold <jobid>    # Place a hold on queued job to prevent it from running

Common SGE settings:

# ------------------------------------------------------------------------------
# sge settings
# ------------------------------------------------------------------------------
#$ -V
#$ -S /bin/bash
#$ -o jid-$JOB_ID-$TASK_ID_jname-$JOB_NAME.log
#$ -j y
#$ -m a \
#$ -M ${USER}@utdallas.edu

Command Example:

qsub /path/to/script.sh --airc_id 3tb1111 --sub 0001 --date 20230101 --ses 3

Testing and Debugging

On cortex, please use either the dev or kenrod partition such as the following commands (adjust parameters as needed; do not run jobs on the login node because this will slow down the server for everyone and, at worse, could crash the server!):

dev partition (preferred)

srun --nodes=1 --ntasks=1 --cpus-per-task=1 --mem=4G --partition=dev --pty /bin/bash

kenrod partition

srun --nodes=1 --ntasks=1 --cpus-per-task=1 --mem=4G --partition=kenrod --pty /bin/bash

Once inside the partition, you can run your script as it is normally executed (e.g., bash /path/to/script.sh)

On ponyo, testing and debugging can be performed by running the script as it is normally executed.

Open OnDemand (OOD)

Open OnDemand allows users to access terminal via a web browser or use programs that require GUI interfaces (e.g., SPM, freeview, or fslview).

Login using the link below using your UTD credentials.

https://ood.cvl.utdallas.edu/

To use terminal:

Clusters > >_cortex Shell Access

To use programs that require GUI interface:

My Interactive Sessions > cortex Desktop > adjust settings as needed and click Launch

Profiles

Consider adding these global lab profiles to your user profile. You can copy and paste either into its respective profile file by using the nano function (e.g., nano ~/.bash_profile)

~/.bash_profile

if [[ ${HOSTNAME} =~ 'cvlkrcompute' ]]; then
  export root_dir='/cvl/kenrod'
else
  export root_dir='/mfs/cvl/groups/kenrod'
fi
source ${root_dir}/server/profiles/global/bash_profile.sh

~/.Rprofile

if (Sys.info()[['nodename']] == 'cvlkrcompute1.utdallas.edu') {
        root_dir <- '/cvl/kenrod'
} else {
        root_dir <- '/mfs/cvl/groups/kenrod'
}
source(sprintf('%s/server/profiles/global/r_profile.R', root_dir))

Technical Information

hostname	alias	cores	memory	cortex node	gpu cores
cvlkrcompute1	ponyo	32	132GB
cvlkrcompute2	totoro	32	132GB	compute-16
cvlkrcompute3	kiki	12	65GB	compute-02	4992