You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Numaflow should have the ability to JOIN during the following scenarios.
JOIN by Key by Window
JOIN by Reduce and using Watermark to close the book. The output will be Keys with Value as an iterator. Please note that the JOIN could be from more than Reduce vertices or consists of one or more Map and Reduce vertices.
JOIN on Map/Sink
A pure non-join write to Sink or Map without really joining by any key.
Multiple Sources
This could be based on the previous two use-cases.
Cyclic Write
We can write back to itself or a previous vertex, but Watermark cannot retrogress, so this can only happen on a pipeline with only Map, and the data written back will implicitly be delayed.
Tasks
A design doc which identifies all the changes needed and challenges
Controller changes to be able to provision join vertex topologies
Watermark changes to support multi-bucket [Map/reduce]
Numaflow should have the ability to JOIN during the following scenarios.
JOIN by Key by Window
JOIN by Reduce and using Watermark to close the book. The output will be Keys with Value as an iterator. Please note that the JOIN could be from more than Reduce vertices or consists of one or more Map and Reduce vertices.
JOIN on Map/Sink
A pure non-join write to Sink or Map without really joining by any key.
Multiple Sources
This could be based on the previous two use-cases.
Cyclic Write
We can write back to itself or a previous vertex, but Watermark cannot retrogress, so this can only happen on a pipeline with only Map, and the data written back will implicitly be delayed.
Tasks
Tasks
The text was updated successfully, but these errors were encountered: