QMUL HPC Usage Policy

1.1 HPC Systems

The HPC system is used by a large number of users, although there are some enforced limits to ensure fair usage, we rely on individual users being aware of the potential impact of their own activities on other users.

Please practice good cluster etiquette by following the policy below and ensuring your jobs are not affecting other users doing this will avoid the need for more draconian technical measures. All usage must also comply with the requirements in Legal Notices

  1. All jobs must be run via the queuing system. It is permissible to log directly into the compute nodes only to check on the state of a job but actual compute work must not be started in this way.

  2. Intensive tasks must not be run on the head/login nodes; these should be reserved for editing code and job scripts, submitting jobs and managing the queuing system.

  3. Compilation should be done in a qlogin session, this is to reduce load on head/login nodes.

  4. Any analysis work must be carried out on your desktop machine or one of the nodes via a qlogin, if more resources are required.

  5. Arrays should be used where appropriate when submitting a large number of jobs. Array jobs with high network or storage usage must use an appropriate Task Concurrency setting to ensure there is no performance impact on the cluster. For information on how to use arrays see the Array Jobs page.

  6. Job submissions should specify the amount of time required rather than using the default. Accurate time requests improve scheduling and result in jobs running sooner.

  7. Job submissions must only request required resources, waste of cluster resources by poorly scheduled jobs will result in your jobs being held or deleted.

  8. Job submissions must only use their allocated resources, using additional resources not allocated by the scheduler will result in the job being automatically or manually deleted.

  9. GPU jobs must only access the GPUs assigned to them via the scheduler as designated by the $SGE_HGR_gpu environment variable.

  10. Use of the cluster is granted for work and research purposes only and resources must not be consumed for personal use.

  11. Jobs or processes that are having an adverse effect on the cluster may be deleted / killed / held as required by ITS-Research with no prior notification. All efforts will be made to inform users once initial issues have been resolved so amendments can be made.

1.2 Access to restricted nodes

  1. Access to restricted nodes will only be granted with permission from the node owner.

  2. Access to GPU nodes will only be granted after an examination of the code and job script, this is to ensure correct usage of the nodes and to avoid wasted resources or conflicted jobs.

1.3 Storage

  1. The "scratch" filesystems are not backed up and data may be deleted from them at any time without prior warning.

  2. Access to Group / Project / Lab shares will only be granted with permission from the share owner.

  3. Sensitive and/or identifiable data must not be stored or processed on the HPC cluster. All data must be anonymised before uploading to the cluster for analysis.

  4. All files / data stored must comply with the Legal Requirements.

1.4 Ephemeral Databases

  1. Databases must only exist for the lifetime of the job, persistent databases will be removed without prior warning.

  2. If space usage increases beyond initial estimates users must contact ITS Research so arrangements can be made to ensure conflict with other usage is avoided.

2 User Accounts

  1. Accounts, passwords and keys must not be shared with anyone and must be kept secure. Failure to comply may result in account suspension without prior warning.

3 Citing Apocrita in your research

Published research should reference this facility using:

This research utilised Queen Mary's Apocrita HPC facility, supported by QMUL Research-IT. http://doi.org/10.5281/zenodo.438045

4 Github Enterprise

  1. You must not store large binary files in repositories.
  2. All files / data stored must comply with the Legal Requirements.

5.1 Access

Access to any data, program or other information on Queen Mary University of London (QMUL) systems must be authorised by QMUL. It is a criminal offence to obtain unauthorised access or make unauthorised modifications to this computer system. QMUL may refer any unauthorised access to appropriate authorities. Unauthorised users may be subject to criminal prosecution and/or civil action and/or QMUL disciplinary proceedings.

5.2 Use

All usage of the HPC systems is also subject to the QMUL IT Regulations.

Users Must NOT process, publish, create, store, download, distribute or transmit material or data that is:

  • Prohibited by UK Law
  • Discriminatory or defamatory
  • Harassing or threatening
  • Derogatory to any individual or group
  • Obscene or pornographic
  • Engaged in any purpose that is illegal or contrary to QMUL policy or business interests
  • Likely to bring QMUL into disrepute

5.3 Monitoring

Queen Mary University of London monitors its IT Systems for the purposes of Operational Integrity, Policy Enforcement and Crime Prevention.