Grid Interactive Jobs
You can schedule an interactive job on the Grid using the FastX service which allows you to run a graphical interface to software such as Rstudio, matlab, Xstata, SAS, and more. It provides an experience similar to working on a desktop or laptop. We recommend you use the FASTX interface for a more responsive graphical interface.
Video Tutorial: Run interactive Stern Grid jobs and singularity containers
Running from a web browser
If you are not on campus, make sure you are logged in to the NYU VPN.
Point your browser to fastx.scrc.nyu.edu:3300 and login using your Stern credentials.
Start a Linux GNOME desktop session; Click PLUS sign → GNOME → Launch.
A Linux desktop opens. Click Activities → Terminal
Step 1 – Running an interactive job on the Stern Grid
Choose the slurm job parameters. Here are the most frequently used options.
| Resource | Description |
|---|---|
--time=hh:mm:ss | wall clock time |
--mem=XXG | maximum memory in GB |
--cpus-per-task=X | number of cpus requested per task |
--nodes=X | number of requested compute nodes |
--gres=gpu:type:N | type and number of requested GPUs where type = v100 or p6000 N = 1 for V100 or N = 2 for P6000 |
Example – start an interactive job
$ srun --pty --mem=12gb --time=01:00:00 --cpus-per-task=2 --nodes=1 --pty /bin/bash
In this example, we are running on the Stern Grid node c4-24a and we have requested 12GB of memory, 2 hours of run time, 2 CPUs per task and 1 Grid node.
To see a list of all the Stern Grid nodes, use the command sinfo.
NOTE: Do not run your jobs directly on the fastx server.
The fastx server is for handling graphical rendering
and not for job processing.
Example – run xstata
As with a batch job, you must choose the type and version of software you want to run. We use Modules to make this selection.
