Skip to main content

Resource Pools & Classes

Compute resources available to Renku users are determined by the resources pools the user has access to. To specify the resources for a project’s launcher, you select a resource class from any of the resource pools available to you. The same resource pools and classes apply to interactive Sessions and background Jobs.

Resource Pools​

A resource pool consists of a total resource quota assigned to a set of users who are allowed to use resources from that quota with culling times for sessions.

  • Resource quota: For example, a resource pool may be configured for a total of 160 CPUs, 1024 GB RAM, and 16 GPUs.
  • Users: There is no limit to how many users can be added to a resource pool. And conversely, a user may be granted access to any number of resource pools. All resource pools the user has access to show up when defining or modifying the default resource class for a launcher.
  • Culling times: The maximum idle time for a resource pool(s) can be configured by the administrator. For example, in the Public resource pool (the default resource pool available to all RenkuLab users), interactive sessions auto-pause after 2 hours of idle time, and shut down after 2 weeks. These thresholds may be customized to be longer or shorter. Jobs are not paused for idle time in the same way as sessions, instead completed and errored jobs are removed from the interface after a retention period (see Job).

Resource Classes​

A resource pool contains a set of resource classes. A resource class is a resource configuration that a user can select for their RenkuLab session or job.

For example, a resource class may be:

  • 12 CPUs, 80 GB RAM, 2 GPUs, 50 GB default disk, 500 GB max disk

As users run sessions and jobs, their compute resources (the resource class in use) come out of the quota on the resource pool. If the resource pool’s quota has been reached, users won’t be able to start more sessions or jobs in that resource pool (but they may use other resource pools they have access to).

ResourceClass For custom resource pools granted to users, you may customize the resource classes in your resource pool.

The Public Resource Pool & Classes​

On RenkuLab, the following resources classes are publicly available (in the “public resource pool”):

ResourcePool

Request a Custom Resource Pool​

Do you need more resources than are available in RenkuLab’s public resource classes? Contact us! We can configure a custom resource pool for your needs upon demand. We apply flat, transparent pricing, passing down infra rates.