[
https://issues.apache.org/jira/browse/IGNITE-22082?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alexander Lapin updated IGNITE-22082:
-------------------------------------
Description:
https://issues.apache.org/jira/browse/IGNITE-22054 adjusted lock behaviour
eliminating the race. Unfortunately we've faced performance degradation after
that.
LockManagerBenchmark.
Was
{code:java}
Benchmark (concTxns) Mode Cnt Score Error Units
LockManagerBenchmark.lockCommit 200 avgt 66.696 us/op
{code}
Now
{code:java}
Benchmark (concTxns) Mode Cnt Score Error Units
LockManagerBenchmark.lockCommit 200 avgt 74.360 us/op
{code}
That might be reasonable, because race was eliminated, however it's required to
add multithreaded benchmark and precisely explain why we have the degradation.
That may lead to locks logic adjustment.
> Investigate HeapLockManager locks performance
> ---------------------------------------------
>
> Key: IGNITE-22082
> URL: https://issues.apache.org/jira/browse/IGNITE-22082
> Project: Ignite
> Issue Type: Task
> Reporter: Alexander Lapin
> Priority: Major
>
> https://issues.apache.org/jira/browse/IGNITE-22054 adjusted lock behaviour
> eliminating the race. Unfortunately we've faced performance degradation after
> that.
> LockManagerBenchmark.
> Was
> {code:java}
> Benchmark (concTxns) Mode Cnt Score Error Units
> LockManagerBenchmark.lockCommit 200 avgt 66.696 us/op
> {code}
> Now
> {code:java}
> Benchmark (concTxns) Mode Cnt Score Error Units
> LockManagerBenchmark.lockCommit 200 avgt 74.360 us/op
> {code}
> That might be reasonable, because race was eliminated, however it's required
> to add multithreaded benchmark and precisely explain why we have the
> degradation. That may lead to locks logic adjustment.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
