Resources

A PERSEUS resource can either be requested by a project or granted to it. PERSEUS distinguishes between two types of resources:

• cumulative resources: resources whose usage is accumulated over time (e.g. CPU core hours or GPU hours)
• snapshot resources: resources whose usage can only be determined based on the last measured data point (e.g. storage used or number of files)

We highly recommend saving resource values using the smallest unit possible, i.e. bytes for storage or single CPU core hours for compute time. This way, you have the most flexibility when working with these values. To make the values human-readable, you can assign a display unit as well as a factor:

• display unit: when a value of this resource is displayed, it will append this unit after the value (i.e. TiB for storage or million for CPU core hours)
• display unit factor: when a value of this resource is displayed, it will be divided by this factor (i.e. 1024^4 for unit TiB or 1,000,000 for unit million)

To group resources and differentiate them, PERSEUS requires that every resource be part of a cluster. This helps to organize them and allows the same resources to be used differently in different HPC systems.

Resource overwrites

Sometimes it can be necessary to temporarily overwrite the available resources and their configuration, for example when

• a compute project should get a higher priority because of paper deadlines,
• a compute project requires temporary access to another partition,
• the available resource value needs to change for a specific time period.

For that reason, PERSEUS allows resource overwrites. The following overwrite types are available:

• Enable partition
• Disable partition
• Set priority
• Change value

Every resource overwrite is limited to a specific time span so that it can be planned in advance and have a predefined end.