Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
Description
Existing Storm metrics track tuple counts, but don't account for the size of the tuple payload. If a task always gets large tuples, it can be slower than a task that gets small tuples without showing any reason why. It would be good to track the byte counts as well so data skew can be observed directly.
Attachments
Issue Links
- links to