Skip to content

Acceptable Use Policy and Guidelines

Overview of the Acceptable Use Policy

The High Performance Computing Center (HPCC) and all of its associated computational resources are covered by the MSU Acceptable Use Policy (AUP). This policy covers all IT resources on campus including "institutional electronic business systems, computing services, networks, databases, and other resources." All Users of the HPCC should familiarize themselves with the AUP and are expected to abide by it.

In particular, please be aware that:

  • The HPCC, as with all other IT resources, is "provided for university-related purposes, including support for the university’s teaching, research, and public service missions, its administrative functions, and student and campus life activities"

  • "Users are responsible for informing themselves of any university policies"

  • "Users are expected to respect privacy of other users"

  • "Users should limit their use of MSU IT resources accordingly and must abide by any limits MSU places on the use of its IT resources"

  • "No user may use any IT resource in a manner which interferes unreasonably with the activities of the university or of other users"

Additionally, use of the HPCC falls under federal rules governing research data. Users of the HPCC should be aware that:

  • Unauthorized use of the HPCC is prohibited
  • Unauthorized use is subject to criminal and civil penalties
  • HPCC usage may be monitored or recorded, and is subject to audit
  • Use of the HPCC affirms consent to monitoring and recording
  • Use of the HPCC must align with requirements in the contact form and statements agreed to on the form regarding sensitive data

As the steward of the HPCC, ICER policies and documentation provide guidelines for how the AUP applies to the system. Documentation about HPCC is available online and users may contact ICER staff with questions at our contact page.

The following pages and policies are particularly important in light of the AUP:

  • Specifications of system hardware can be found on the Cluster Resources page

  • Policies concerning running jobs on HPCC can be found on the Job Policies page

  • Instructions on obtaining an account can be found on our HPCC access page

    • Each user must have their own account as sharing accounts may lead to intentional or unintentional exposure of a user's private information

    • All user accounts and data are subject to ICER's User and Data Policies

  • Users must not monitor system function or the behavior of other users themselves. Concerns should be reported to ICER staff at our contact page.

  • ICER will intervene in cases where the staff determine activity 'unreasonably interferes' with the normal function of the HPCC

    • We understand that most cases for unreasonable use are unintentional, but we may still cancel jobs or remove access when we determine it will unfairly impact other users or affect system stability

Recommendations from ICER

ICER recommends that users consult our documentation website whenever they are unfamiliar with or unsure about any aspect of the HPCC. ICER also offers in-person and asynchronous training options for subjects such as HPCC fundamentals and common programming languages.

We would also like to highlight the following topics for the benefit of our users:

  • Please familiarize yourself with how to use the SLURM schedueler and our guide on requesting SLURM Resources. This will help you run jobs efficiently within the limitations of our job policies and hardware

  • When working with collaborators in a research space, please see the Research Space page for information on how permissions and the quota work in this space

  • When possible, make use of services provided by HPCC rather than coding your own. For example:

    • ICER provides graphical interfaces for RStudio and Jupyter through our OnDemand Service

    • Many software packages can be installed locally through Conda

    • Using the SLURM scheduler as described is more efficient than trying to find and tune your jobs for open nodes. If your jobs are queueing slowly, it often because of priority or requesting more run time than is needed.

  • When working with software on HPCC, familiarize yourself with how the software works, what resources it requires, and where it transmits data to. You are responsible for the behavior of software you run on HPCC. This is particularly important with regard to emerging large language models (LLMs) and AI tools. Please refer to MSU's Guidelines on Artificial Intelligence.

Working within our existing recommendations and tools tends to minimize friction which could lead to behaviors that interfere with the function of HPCC.

Examples of behavior that interferes with HPCC

Determining what type of behavior interferes with the function of HPCC is at the discretion of ICER. Given the complexity of the system and the variety of research projects we seek to support, our staff needs the flexibility to intervene with new problems as they occur. Therefore, it is not feasible to provide an exhaustive list of behaviors which cause unreasonable interference.

However, below we would like to highlight certain behaviors in the past which have required intervention by ICER:

  • Running computationally-intensive code on the gateway nodes: the gateway nodes exist to manage user access to HPCC and occupying resources on these nodes can slow down logging in system wide. Please use the dev nodes to test your code.

  • Attempting to circumvent limits on jobs on the dev node: the dev nodes are a common resource for all users to test software and have resource limits for running jobs

  • Abusing the SLURM scheduler: the SLURM scheduler exists to fairly distribute resources among users. Users should not (1) game the schedulers by asking for more resources than a job needs nor (2) submit jobs with excessive resources requested to ensure they complete. If you require assistance determining what resources you jobs needs, please use our resource request guide.

  • Attempted access of systems and folders you should not have access to: all users are expected to respect the privacy of other users and it is a violation of policy to intentionally seek access to directories that have not be shared with them

  • Monitoring other user's activity: if a user believes the activity of another user is causing problems with HPCC or violating the AUP, they should report this to ICER using the contact page. Monitoring, collecting data about, or contacting another user directly about their activity on HPCC is a violation of their privacy.

  • Overuse of system resources

    • Job Scheduler: excessively calling SLURM commands (squeue, sacct, sinfo) on a short interval
    • File System: excessively writing to the same file, such as from multiple jobs and/or multiple times a second
    • On node resources: writing directly to /tmp. Please see the page on Local File Systems for how to make use of /tmp.

Enforcement

As stated in the AUP:

"Use of MSU IT resources is a privilege and not a right. A User’s access to MSU IT resources may be limited, suspended, or terminated if that User violates this Policy."

ICER staff will interrupt jobs and suspend accounts which we determine are interfering with normal HPCC function. We understand that such behaviors are often unintentional and, as such, ICER staff will attempt to reach out to users to resolve issues after such intervention. However, our first priority is to ensure that the HPCC remains functional for all users.

Users who repeatedly do not comply with warnings or who act with malicious intent may permanently lose privileges on HPCC and may be referred for disciplinary action under the AUP. Furthermore, PIs are responsible for the conduct of their sponsored users, and entire project groups may be penalized for continued abuse of the system.