Github user revans2 commented on the issue:
https://github.com/apache/storm/pull/2261
From a code perspective the changes look fine to me. It would be nice to
possibly clean up the load aware grouping Interface some.
From a performance perspective it is a bit of a wash from what I am seeing.
I ran ThroughputVsLatency with 1 spout, 1 splitter, and 1 counter. I also
did the same with 2 splitters, so there was a choice to be made.
From a cost perspective I saw that for the 1 splitter case everything
appeared to be within the noise range. For the 2 splitter case I saw the new
version taking up about 5% more CPU than the old version, so I would like to
explore this a bit more.
Similarly for latency at the mean, 99%ile and 99.9%ile latency
measurements. The original one was slightly better, around 5%, when there were
2 splitters.
I do want to explore these a bit more, because it seams counter the
benchmarks that others have run.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
