Details
-
Sub-task
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
Description
Preferred locations based on state and inputs are scattered into multiple components, which makes it harder to reasoning the calculation and complicates those hosting components.
We can introduce a PreferredLocationsRetriever to be used by ExecutionSlotAllocator which returns preferred locations of an execution vertex and hides the details of the calculation from other components.
Attachments
Issue Links
- links to