Skip to main content

Celonis Product Documentation

Managing machine learning resources and consumption

Active machine learning workbenches (and the notebooks they contain) consume resources, notably memory, CPU, and data storage. These resources count towards your allocated Celonis Platform team resources, so attention must be paid to how much your existing machine learning workbenches are consuming.

Note

These features are limited to admins of Celonis Platform teams who have upgraded their machine learning license to include dedicated resource management.

As an admin, you can manage your machine learning resources and consumption levels by clicking Configuration:

ml_resource_configuration.png

You then have the following resource management options:

Auto shutdown of machine learning workbenches

By default, unless machine learning workbenches are marked as productive, they are shut down after 12 hours of inactivity. However, you can enable and configure an automatic shutdown of any machine learning workbenches which have not been used for a defined period of time. Once they're shut down, workbenches no longer consume resources. However all the information and configuration details are saved.

Long running calculations that are executed from the Jupyter UI might be canceled if they are running for longer than the specified time. Scheduled executions are not affected by this and are only restricted by the timeout settings in the scheduler settings.

You can also manually shut down a workbench by clicking Options - Shutdown.

manually_shutdown.png

Create additional machine learning resource pools

If your Celonis Platform team has resources available, you can create and assign additional resource pools for your machine learning workbenches. Resource pools help you to allocate and then identify where your resources are being consumed within your Celonis Platform.

To create additional resource pools, click Add Pool.

add_pool.png

Assign workspaces to machine learning resource pools

Workspaces allow you to further organize your machine learning workbenches, a useful feature if you want to group commonly used workbenches together. You can also then simultaneously assign permissions to multiple workbenches.

To create a workspace, click Apps - New Workspace:

new_workspaces.png

You can then assign workspaces to your machine learning resource pools by clicking Configuration - Assign Workspaces.

configuration_assign_workspaces.png