Skip to main content

Celonis Product Documentation

Resource Management for Dedicated Resource Customers

Note

This feature is only available for teams with an upgraded machine learning license with dedicated increased resources (Memory, CPUs, Storage and in some cases GPUs).

Teams with dedicated resources can easily manage and distribute these resources between different Machine learning workbenches, by easily updating the memory and CPU limits as well as start and stop the workbenches to free up the resources as required. Only admins of the team are able to manage these resources between all the ML Workbenches and see an overview of all the resources available.

  1. While creating a workbench you will be able to also select the memory, cpu, storage and potentially the gpu (where available) limit for the workbench as required by a specific project you are creating. You can also mark a workbench as productive. Marking a workbench as productive will mean that the resources are blocked for this workbench and will always be available, that is, this workbench will never be shut down automatically in case of longer inactivity. We recommend to mark the workbenches productive if they have scheduled executions planned.

    41193477.png
  2. Under the configuration tab in the side menu, you can see the overview of the status of all the resources as well as the auto shutdown setting. This shows how many total resources are allocated to the team and how much is available out of those resources. This also shows an overview of which ML workbench is occupying how much memory, cpu and storage. You can also see which of the ML workbenches are marked as productive here.

    Note

    This view is only available for admins of the team.

    41193478.png
  3. 3. An admin can then start or shutdown any ML workbench from the configuration page by clicking on the 3 dots menu next to each row. You can also edit the resources of the workbenches from here using the Edit option under the same menu.

    41193486.png
  4. Auto Shutdown time of any non-productive ML workbench is 12 hours by default in case of inactivity. This can be updated according to the needs (increase or decrease) by enabling the auto shutdown toggle from the configuration page. You need to put the auto shutdown time period in hours in the respective filed and hit save button. Auto shutdown help free up resources that are unnecessarily occupied by the idle ML workbenches. An ML workbench that is shut down does not lose any information or data and only frees up the resources occupied by it. It can be started again any time.

    41193479.png