Guidelines and policies#

The following describes the general guidelines for using the Esrum cluster. Failure to follow these guidelines may result in your tasks being terminated and may even result in your access to the cluster being revoked.

Please also see the general UCPH resources for handling GDPR sensitive data.

Basic security#

  • Your account on Esrum is strictly personal and must not be shared.

  • Never leave your computer unsecured while logged onto the cluster. Your computer must be locked or turned off whenever you leave it.

Data storage#

  • GDPR protected data must be stored in audited folders. These can be recognized by the -AUDIT suffix, for example /projects/name-AUDIT or /datasets/name-AUDIT.

  • Directory names and file names must not contain GDPR protected data or other confidential information, even if placed in an audited folder. This is because common operations expose this information to all users of the cluster. For the same reason, do not include such information in command-line arguments.

  • Data must not be copied out of audited /datasets or /projects folders without permission from the relevant data controller. Instead, use symbolic links if you need the data to be located in a more convenient location.

  • Data must not leave the cluster without permission from the relevant data controllers.

  • Do not store data or other files related to projects in your home folder. Your home folder is accessible only to you, unless you have provided written consent to UCPH-IT.

See Data storage on Esrum for more information.

Use of Generative AI and LLMs#

Use of Generative AI / Large Language Models (LLMs) is permitted in accordance with the general guidelines and rules for using ChatGPT and similar technologies at UCPH: In Danish and in English.

Briefly,

  • It is your responsibility that no sensitive information is transmitted to external systems.

  • If possible, run the models locally on Esrum using our GPU nodes.

Running jobs#

  • Do not run big jobs on the head node (esrumhead01fl), as doing so may impact the ability of everyone to use the cluster. However, we do permit small jobs on the head node, meaning a few cores in total and modest memory usage.

  • Remember to be considerate to other users. For example, limit the number of jobs you are running simultaneously, so that others users can also run their jobs

  • Please remember to close interactive shells, notebooks, containers, and other processes that you have started via Slurm or the container system. Resources that you have reserved are not made available for other users until your tasks have finished.

See the Running jobs using Slurm page for more information about how to run your tasks on the cluster.

Access to protected data by staff#

The Data Analytics Platform (DAP) may require access to protected data in order to provide services requested of the platform. This may include maintenance, data processing, formatting, storage, indexing, curation, and more.

When this need arises, DAP will either

  • grant access to the relevant staff member(s), for an appropriate time period, after notifying the data owner. This applies to datasets and projects directly administrated by DAP.

  • request access to the datasets and/or projects, using the Identity Management system. This applies to datasets and projects not administrated by DAP.

A formal data access request will also be made in cases where DAP staff member participate in the scientific exploitation of protected data, beyond that required to fulfill technical or service functions.

Immediate intervention under the supervision of a user with appropriate access to protected data, for example while providing direct support, is not considered establishing privileged access.

Acknowledgements#

If you make use of the Esrum HPC cluster, human cohorts managed by Data Analytics, or otherwise receive support from the platform, then we ask that you kindly acknowledge the platform in any resulting publications, presentations, or other scientific output. For example via an acknowledgement such as,

HPC facilities and technical expertise was provided by the Data Analytics Platform at the Novo Nordisk Foundation Center for Basic Metabolic Research.