Data Scientists have access to a wide range of ML Runtimes, they can use different versions of Python, R, and Scala kernels with the Workbench or the JupyterLab editors and can benefit from extended capabilities like GPU acceleration. Cloudera continues to release new versions of ML Runtimes, while customers can also register custom ones they build to solve their specific business use case. With the continuous additions, the available options can grow large, and some will become irrelevant or outdated.
CML now supports administrators to disable Runtime variants or specific versions. For example, they can decide to disable all of the Python 3.6 Runtimes as the python kernel is officially EOL and won't receive any further security and bugfix patches. Usage of these runtimes can be considered a security risk that administrators now can solve. Once administrators disable an ML Runtime, data scientists won't be able to use them for development, and existing workloads configured with them will also fail to start.
