Job Input

When submitting a job to the TACC HPC systems using SLURM – whether through the DesignSafe portal, TACC web portal, or command line – you must provide a set of input parameters. These parameters control how your job runs, how many resources it uses, and where input/output files are stored.

Parameters for All HPC Jobs

These input fields are required for all types of jobs, regardless of what software or simulation tool you’re running.

Allocation

Specify the TACC allocation account to charge for this job. A TACC allocation is a grant of computing resources – such as CPU/GPU time, storage space, or access to specific systems – on TACC’s supercomputers.

• This is often associated with a funded research project or proposal.

• Choose from the list of allocations available to your user account.

• If unsure, consult your PI or project manager.

• Check on your allocations by accessing your TACC User Portal directly.

Queue

Select the queue (also known as a partition) where the job should run.

• Different queues support different workloads .

• Refer to the Queue documentation to determine the best choice based on job duration and resources.

Maximum Job Runtime

The maximum wall-clock time your job should run, in .

• Your job will be terminated automatically if it exceeds this time.

• Shorter runtimes typically result in faster scheduling.

• Platform-specific limits apply (e.g., usually 48 hours for the normal queue).

• When submitting job runtime in a SLURM script, it is specified in HH:MM:SS format Example: 02:00:00 for 2 hours.

• Note: The time required to transfer your data files to and from scratch is included in your job run time. –> Make sure you leave enough time for data transfer, otherwise you will have to transfer your results manually via ssh from scratch to your data.

Node Count

The number of compute nodes your job requires.

• Each node is an independent machine.

• You are billed per node, so only request what you need.

• Default is often 2 nodes, but smaller jobs may only need 1.

Cores per Node

The number of CPU cores (processors) per node to use.

• Total processes = Node Count × Cores per Node.

• More cores = higher parallelism, but also higher memory use.

• For memory-heavy jobs, consider using fewer cores per node to allocate more memory per core.

Note: Different TACC systems have different cores per node.

Job Name

A custom name for your job.

• Helps you identify jobs in the SLURM job queue and job output.

• This name will also appear in output directories and logs.

• Useful when submitting multiple jobs.

Archive System

Specifies the storage system where job results will be saved after execution.

• Defaults to designsafe.storage.default (MyData) or another predefined system.

• Best left at the default value unless you have specific storage needs.

Archive Directory

The folder where output files will be copied after the job completes.

• This is usually inside your MyData space – look for the MyData/tapis-jobs-archive folder. There the jobs are stored in nested folders where they are first grouped in a folder for each day. Within that folder you will find your job’s folder, which should include the job name.

• You can specify a custom location if needed, but the default is recommended.

• Output directories are automatically named using the job name and timestamp.

Additional Parameters

These may be Required by some Applications

Some DesignSafe applications automatically generate the SLURM job script for you. Others — like the OpenSees Web-Portal app — may also prompt you for additional parameters to handle extra tasks.

For example, the OpenSees Web-Portal app not only prepares the SLURM input but also copies your input files to the scratch directory on the HPC system, runs your analysis there (which is faster due to high-performance scratch storage, taking full advantage of the HPC architecture.), and then retrieves the results back to your DesignSafe workspace once the job completes.

Input Directory

This directory contains all input files for your job.

• Includes your main script and any secondary files.

• This directory is copied to the HPC system at job start.

• Once the job finishes, the entire folder (including outputs) is copied to the output/archive location.

Main Script

The name of the main input file that the application, such as OpenSees, should execute.

• Must be located inside the Input Directory.

• This file is passed to the proarm as a command-line argument.

• All other script calls must originate from this file.

Tip

Use relative paths inside your input script so they still work after the input folder is copied.

Summary Table

Parameter	Applies To	Description
– Job Parameters –
Allocation	All jobs	Select which TACC allocation/account to charge.
Queue	All jobs	HPC queue to run in (e.g., skx,skx-dev).
Max Runtime	All jobs	Maximum wall-clock runtime for the job.
Node Count	All jobs	Number of compute nodes to request.
Cores per Node	All jobs	Number of processors per node.
Job Name	All jobs	Descriptive name for the job (for tracking/logging).
Archive System	All jobs	Storage system for saving job results.
Archive Directory	All jobs	Folder to save job output files.
– OpenSees-Analysis Parameters –
Input Directory	Some jobs	Folder with scripts and input files.
Main Script	Some jobs	Main input script passed to the application at runtime.