Skip to Content

Division of Information Technology

Research Computing


Research Computing requires all users of its research computing resources to submit jobs through the SLURM queuing systems. Do not attempt to bypass or hinder these systems. Documentation for the queuing systems on each major computing resource is available on this web site. Please be as accurate as possible about the resources which your jobs will require, as this will help the system run more efficiently and may help your jobs run more quickly. You must specify the actual number of processor cores that each job will use when submitting your jobs. Failure to do so could result in poor performance and may adversely impact other users' work.

All users of Research Computing resources must also comply with USC’s IT Resource Acceptable Use Policy, UNIV 1.52, Responsible Use of Data, Technology, and User Credentials and USC’s Remote Access to IT Resources Policy HR 1.22, Telecommuting [pdf].

Data about activity on Research Computing resources is routinely logged, archived and analyzed, to provide feedback about system performance, resource and software utilization, and better optimize the resources. This includes, but not limited to, environment modules loaded, applications run, hardware performance counters, disk storage consumed, and the contents of batch job scripts.

Front-end or login nodes for Research Computing research clusters and other shared computational resources are available only for use in job preparation and submission, file editing and transfers, compiling and editing code, and monitoring jobs. You should never use the login nodes to run computationally or memory-intensive applications or otherwise do computational work. Use of memory and CPU on these nodes should be minimal, as all users of the cluster or resource must share these systems. You should submit any work to a queue for processing on the compute nodes of the cluster.

Any processes using significant processor time or memory on a login node will be terminated without warning. You will receive an email notifying that your processes have been stopped.

If you believe a process or application that meets acceptable use was incorrectly terminated, please let us know at and Research Computing staff will review the application.

Research Computing operates a number of SLURM Job Queues that allow open access to nodes shared by the university community and partners having exclusive access to the nodes they have purchased. The  queues offer priority access to a portion of the system as paid for by the researcher, but also offer access to the default queue the USC community who have not contributed compute nodes, whenever it would otherwise not be busy.

To facilitate this, all community clusters run a scheduling system that offers multiple queues for job submission. At least one queue is created for the partner group's priority access. This is restricted to the number of nodes and cores to which access was purchased, allows for very long jobs, and is usually named after the group or faculty member.

Users who have not contributed nodes may send jobs to the "community" queue. This queue is open to anyone, but all community users share resources with each other.

Runtime of jobs sent to community queues may take much longer than partner runtimes using their dedicated nodes and queues depending on the total shared load of all community users.

Each Research Computing research resource has different account eligibility requirements. For details, please refer to the documentation for the specific resource that interests you. Generally, resources will allow users from one or more of the following groups, though not all groups on all resources:

  • USC faculty and staff who purchased nodes in the resource and their designees
  • USC faculty, staff, and students
  • People directly affiliated with USC and/or working closely with faculty or staff at USC
  • Users of a research project of which USC is a member (XSEDE, etc.)

Research Computing may deactivate accounts within 30 days of termination of the relationship which authorized them. Reasons for termination include: leaving USC, leaving a project at USC, or ending an active partnership with a USC researcher.

Users must download any data or files they wish to keep prior to deactivation.

Research Computing reserves the right to terminate access to an HPC account at any time under any circumstances without warning.

Research Computing provides shared community nodes at no charge and partner nodes with priority given to specific faculty or research groups.  These are several policies regarding access to these different resources.

The Community Queue, shared by USC students, faculty, and staff, can access all community compute, GPU and Big Data nodes including those purchased by partners when not being used by a partner and their designated users.

  • Community users must complete for shared resources with all other community users that may result in longer runs times when compared to Partners who purchase their own dedicated hardware.
  • Researchers collaborating with faculty who have purchased nodes should request being added to the Partner’s queue if appropriate. The Partner queues provide dedicated resources and less competition for resources compared to the shared Community Queues.
  • Research Computing can provide Bolden accounts for workshops or teaching. These are considered limited term accounts for the duration of the workshop. Accounts may remain active for a period of time after each workshop to allow further learning. To continue using Planck for research or further self-study, please request an account as follows.
  • Research Computing asks for the name of a faculty member for each account so that we may accurately attribute usage to PIs, faculty, and departments. There is no charge for usage of the cluster. We only use this information for accounting purposes.

Partner Queues are not generally available to all USC students, faculty, and staff. All nodes associated with partner queues are given priority to users designated by the partner.

  • Researchers collaborating with faculty who have purchased nodes are asked to use those partner queues instead. The partner queues provide more dedicated power than community queues.

Teaching clusters are available to USC instructors from any field whose courses include assignments that could make use of supercomputing.

  • Bolden accounts may be requested for any officially recognized USC course or workshop by the instructor.
  • In compliance with FERPA regulations, it is not necessary to list every student on the request. Simply provide the USC USERID and an optional list of instructors or TAs for that course.
  • Bolden is NOT intended for research purposes, PhD Thesis, MS Thesis, Undergraduate Research or Graduate Research type course registrations. Bolden is intended for instructional purposes only.
  • Research Computing may provide HPC accounts for sponsored workshop programs. These are considered limited term accounts for the duration of the workshop. Accounts may remain active for a period of time after each workshop to allow further learning. To continue using teaching queues for research or further self-study, please request an account as follows.
  • To request an account, use the Access Request When prompted, enter your advisor or PI's name (if you are the faculty member, enter your own name).
  • Research Computing asks for the name of a faculty member for each account so that we may accurately attribute usage to PIs, faculty, and departments. There is no charge for usage, we only use this information for accounting purposes.

Home directory and high-speed scratch storage is generally available to USC students.    A  scratch space of the 300 TB is shared by all users.

Each user is responsible for saving any intermediate files on scratch to their home directories then backing up their home directories external to the cluster.

Research Computing does not provide any commercial software licenses. Installation of commercial software on Research Computing HPC systems must include a valid license and proof of license eligibility. Research Computing will install and maintain the software on the Hyperion cluster upon receipt of the software, licensing and vendor contact information.

Research Computing upholds all USC University policies on intellectual property.

Users of Research Computing research systems authorized by a Principal Investigator (PI) agree to allow access by their PI to all data and files remaining on Research Computing research systems after their account is deactivated.

Generally, Research Computing systems administrators will look at a user's files only if this is deemed absolutely essential to diagnosing a problem. In addition, should this be necessary.  Research Computing does reserve the right to access or disclose any file or data should this be deemed necessary to serve the interests of USC.

Storage Space Limit
 Home Directory  25 GB
 Scratch Space  Managed by Research Computing

Research Computing provides all users of research computing systems with a scratch directory. Research Computing research scratch directories are available for short-term storage of files. There is no backup service for scratch directories, and files not accessed or modified in the last 90 days are subject to removal. In the event of a disk crash or file removal, files in scratch directories are not recoverable.

Users are responsible for saving copies of all important files external to the cluster on a regular basis.

If you need help in transferring your files, contact Research Computing at

Scratch directories are purged on the basis of last access time and content modification time of an individual file. Any file not accessed or had content modified in 90 days will be subject to purge. Changing file metadata, such as file name or permissions, does not protect a file from purging.

It is important to keep in mind that cluster scratch space is for limited-duration, high-performance storage of data for running jobs or workflows. Old data in scratch file systems is occasionally purged to keep the files ystem from being fragmented or filling up. Scratch is intended to be a space in which to run your jobs, and not used as long-term storage of data, applications, or other files.

Please keep in mind that any scratch file system is engineered for capacity and high performance, and are not protected from any kind of data loss by any backup technology. While research computing scratch filesystems are engineered to be fault-tolerant and reliable, some types of failures can result in data loss.

Making frequent copies of your files will minimize work required when these files eventually become subject to purge, as well as protect your work in the unlikely event of a scratch system failure.

Please send email to if you have questions or need assistance.


Challenge the conventional. Create the exceptional. No Limits.