Policies and service conditions¶

Overview¶

UCT HPC is a shared institutional resource designed to support research.

Use of the system is governed by policies and service conditions to ensure:

fair access across users and groups
system stability and performance
appropriate use of institutional infrastructure

All users are expected to understand and follow these conditions.

Core principles¶

Use of HPC is based on the following principles:

shared resource — all users must act in a way that does not negatively impact others
research-focused use — the system is intended for legitimate research activities
responsible usage — users are accountable for how they use compute, storage, and software
reproducibility and traceability — workflows should be structured and documented

Appropriate use¶

HPC resources may be used for:

academic research
data analysis and modelling
simulation and computation
development of research software

HPC should not be used for:

personal or non-research purposes
unrelated commercial activities
activities that violate institutional policies or law

The login (head) node is a shared access point, not a compute resource.

Allowed:

connecting to the system
editing files
submitting jobs
lightweight commands

Not allowed:

running computational workloads
running graphical applications
compiling large software packages
long-running or resource-intensive processes

Misuse of the login node may result in:

termination of processes
temporary or permanent restriction of access

Compute usage¶

All computational work must be run via:

batch jobs (scheduled)
interactive jobs (for development or GUI use)

Users must:

request appropriate resources (CPU, memory, time)
avoid over-requesting resources unnecessarily
release resources when no longer needed

Fair usage¶

To ensure equitable access:

jobs may be prioritised or limited by the scheduler
excessive or inefficient use may be restricted
users may be contacted if usage patterns affect others

Good practice includes:

testing jobs on small inputs before scaling
optimising code before large runs
avoiding repeated failed jobs

Storage usage¶

Users are responsible for managing their data.

Expectations:

store only active and relevant research data
remove unnecessary or duplicate files
avoid using HPC storage for long-term archiving (unless designated)

Large or unmanaged datasets may:

impact system performance
be subject to cleanup policies

Software and installation¶

Software environments are shared.

Users must:

use centrally provided software where appropriate
avoid modifying shared environments
install personal packages in user space

Heavy installation or compilation must:

not be performed on the login node
be done via interactive jobs

Security and access¶

Users are responsible for:

keeping credentials secure
not sharing accounts
ensuring appropriate access to data

Do not:

expose HPC services insecurely
bypass authentication mechanisms
run services that create security risks

Data responsibility¶

Users are responsible for:

the integrity of their data
backing up important results where required
complying with data governance and ethical requirements

HPC systems are not guaranteed backup environments unless explicitly stated.

Monitoring and enforcement¶

System usage may be monitored to ensure:

compliance with policies
system stability
fair resource allocation

If misuse is detected, actions may include:

warning or guidance from support staff
termination of running processes
temporary suspension of access
escalation under institutional policies

When to ask for guidance¶

Contact HPC support if you are unsure about:

whether a workflow is appropriate
how to use resources efficiently
large-scale or unusual workloads
software installation requirements

Early engagement helps avoid issues.

Good practice summary¶

use compute nodes, not the login node
run jobs through the scheduler
request appropriate resources
clean up unused data
document and structure workflows
respect that HPC is a shared system

Scheduler and job submission → Reference > HPC
Software and modules → Reference > HPC
Storage and file systems → Reference > HPC
Graphical applications → Reference > HPC
Support → Support