rohangarg opened a new pull request, #12817:
URL: https://github.com/apache/druid/pull/12817
This change fixes the hash calculation logic in RendexvousHasher.
@abhishekagarwal87 pointed that RendezvousHashing based JDBC connection
balancing produces a skew in the distribution of the connections amongst
brokers.
Upon checking further, it was found that the `combineOrdered` call to
generate a composite hash wasn't uniformly distributing the hash value amongst
all the brokers. So, the new logic creates the composite hash using the
`Murmur_128` hasher which leads to a uniform distribution of load. On testing
locally with 8 brokers, the old logic could produce as much as 20% skew between
the lightest and heaviest broker. After this change, the skew is consistently
observed to be < 5%.
The benchmark run also shows improvement in runtime with the new hashing
logic.
```
Before
Benchmark (numIds) Mode Cnt Score Error
Units
RendezvousHasherBenchmark.hash 100000 avgt 30 129676.458 ± 950.211
us/op
After
Benchmark (numIds) Mode Cnt Score Error
Units
RendezvousHasherBenchmark.hash 100000 avgt 30 97226.880 ± 1019.564
us/op
```
This PR has:
- [x] been self-reviewed.
- [ ] using the [concurrency
checklist](https://github.com/apache/druid/blob/master/dev/code-review/concurrency.md)
(Remove this item if the PR doesn't have any relation to concurrency.)
- [ ] added documentation for new or modified features or behaviors.
- [ ] added Javadocs for most classes and all non-trivial methods. Linked
related entities via Javadoc links.
- [ ] added or updated version, license, or notice information in
[licenses.yaml](https://github.com/apache/druid/blob/master/dev/license.md)
- [ ] added comments explaining the "why" and the intent of the code
wherever would not be obvious for an unfamiliar reader.
- [ ] added unit tests or modified existing tests to cover new code paths,
ensuring the threshold for [code
coverage](https://github.com/apache/druid/blob/master/dev/code-review/code-coverage.md)
is met.
- [ ] added integration tests.
- [ ] been tested in a test Druid cluster.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
