A regional emergency-management partnership wants mutual-aid forecasting teams in different Azure Machine Learning workspaces to consume the same registered model-component catalog. What should they establish?

Establishing an Azure ML registry that shares the component catalog across the partner workspaces is correct for a regional emergency-management partnership because mutual-aid forecasting teams need a common, governed catalog of model components, not just a shared Azure ML compute cluster attached to partner workspaces for training. A registry provides identity, versioning, and cross-workspace consumption so each partner Azure ML workspace can pull the same registered components. That supports collaboration while preserving separate operational workspaces.

A shared Azure Container Registry that distributes container images to partner workspaces provides image storage and distribution, not a governed component catalog — partners cannot resolve the same versioned, registered model-component definitions because it lacks component identity and cross-workspace component consumption. A shared Azure ML compute cluster attached to partner workspaces for training provides compute capacity, but it is not a governed component catalog and does not scale regional model sharing. Versioned Azure ML data assets publishing the forecast datasets to partner workspaces share data, not components, and cannot serve as a shared registered component catalog.

Exam caveat: registries are the cross-workspace sharing fabric for components, environments, and models. Distinguishing "share catalog across workspaces" (registry) from "train on a shared Azure ML compute cluster" is a common exam discriminator.

Operational check: stand up or designate the partnership registry, publish the mutual-aid forecasting components, grant partner workspaces pull access, and verify each workspace can resolve the same component version in a pipeline.