Acceptable Use Policy
By logging on to the ICTS HPC cluster you are agreeing to abide by UCT’s policy and rules on Internet and email use as well as the HPC Acceptable Use Policy terms and conditions outlined below;
General HPC terms and conditions
• Computing resources including but not limited to CPUs, RAM, disk space, internet bandwidth and networking are to be used for research purposes only.
• User accounts may not be shared.
• No commercial work is permitted.
• No proxy work, you may not submit jobs on behalf of someone else.
Revocation of access
Any infringement of the above terms, or if the following abuse is noted, may result in your user account being suspended and/or disciplinary action:
• Running jobs or any high load process on the head node.
• Transferring large volumes of data via the head node.
• Viewing or editing extremely large files on the head node.
• Over or under subscribing resources.
• Running serial code as a parallel job.
• Running the screen command on a worker node.
• Causing a file system to run out of space.
• Allowing illegal access to the cluster.
• Downloading non research related files or data.
Core, Memory and Time limits
By default, users will have access to 64 cores (soft limit). This can be boosted to 96 cores (hard limit) or beyond if the cluster is under-utilised or if the usage patterns of a researcher’s jobs can be anticipated. If jobs are queued and the cluster is severely under-used then these jobs will be allowed to run by discretion of the HPC system administrators. The hard and soft limits will be altered over time as hardware and usage patterns change in order to achieve maximum efficiency and usage of the cluster.
If your jobs are known to have a low wall time and\or you make use of the wall time directive then depending on the current state of the cluster your core usage may be increased substantially at the discretion of the administrators. It is also possible to access additional cores by purchasing computational hours.
Users must abide by the ratio of 2GB of memory per allocated core. Usage of the pmem directive should also be observed.
The maximum wall and cpu times are displayed on the dashboard under queue parameters. The default wall time is set to 250 hours. If your jobs need to run longer you may use the walltime directive to increase your walltime up to the allowed maximum.
When your account is deleted all files on /home and /scratch are deleted. It is your responsibility to move data off the HPC file systems when your research projects come to an end.
/home: There is limited disk space on /home and this is shared with other researchers. You are limited to 20GB in /home for scripts or reference data. This volume is protected bu ICTS’s backup system although there is a cost for restores.
/scratch: The /scratch volume provides extremely high speed disk access for transient research data. You will be granted a 100GB quota by default. The /scratch volume is volatile and is not backed up. Users are expected to move data off this volume once computation is completed. Old files will automatically be deleted if the volume starts to run out of space. Using the touch command to modify files in the /scratch volume will result in your user account being revoked. The 100GB quota can be increased on request.
|20TB||Depends on usage|
Long term storage: Users that require reliable long term central storage should contact email@example.com
All user home directories are protected at the file system level. Researchers that request shared data areas must inform ICTS HPC staff as to which HPC users or groups may have access to these areas. ICTS HPC administrators reserve the right to monitor disk space usage and/or file contents.
In order to maintain the stability of the HPC clusters, the ICTS HPC administrators reserve the right to terminate any running job that does not fall within the agreement of good usage, or that threatens the running of other researchers’ jobs.
Data and disaster recovery
You are responsible for your data. Recovery costs due to data loss caused by user action or inaction will incur a cost of R200. Only data that is stored on /home is backed up to tape. Data stored on these tapes will be kept for a maximum period of one year, thereafter the tapes will be recycled. Should you need to keep your data for a longer period, you could save your data on the research data archive repository.
Compilation of software
Compiling of software or installing libraries should not be done on the head node. This should be run as a job. If debugging or interaction is required then you must run an interactive job.
Large volumes of data (>1GB) must be transferred via hexdata.uct.ac.za, not the head node.
Researchers are required to notify us of any submissions, publications or presentations that contains work conducted using the HPC resources. The following form of acknowledgement must also be included in your published work:
Computations were performed using facilities provided by the University of Cape Town’s ICTS High Performance Computing team: hpc.uct.ac.za
If you need to access cluster from outside of UCT you will need to do so by making use of the VPN.