Github user eshioji commented on the pull request:
https://github.com/apache/storm/pull/568#issuecomment-108978724
@revans2 I was able to bring back the performance where it was, if not a
bit higher:
| | 0.9.5-SNAPSHOT | STORM-763 |
|-----|----------------|-----------|
| 10 | 23.8 | 23.9 |
| 1K | 78.6 | 82.5 |
| 2K | 75.6 | 74.5 |
| 10K | 71.1 | 72.9 |
The change brings back the batching of in-between calls. However, it does
it a bit differently; instead of using a background process, it uses Netty's
Channel interest notification callback. This should have an added benefit in
theory, because pending messages are immediately flushed as space becomes
available, rather than potentially waiting up to FLUSH_CHECK_INTERVAL_MS. I
also brought back @miguno 's graceful shutdown. Let me know if this is good.
One question; I noticed that you guys plan to make 0.9.5 the last 0.9.x
release. This PR is against 0.9.x, but should I rather open it against 0.10.x
(or master)?
PS
@miguno No worries at all, I'm glad I can give back, however small it may
be :)
---
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.
---
