[
https://issues.apache.org/jira/browse/HDDS-9377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17923947#comment-17923947
]
Alex Juncevich edited comment on HDDS-9377 at 2/5/25 7:51 AM:
--------------------------------------------------------------
Without a fix, it was a 3000-3200 writing operation (CreateFile and CommitKey).
After fixing it was 3600-3700. Improvement of more than 12%.
was (Author: JIRAUSER306129):
Without a fix, it was a 3000-3200 writing operation (CreateFile and CommitKey).
After fixing it was 3600-3700.
> Call MutableStat.add(long x) method as a runnable task in a separate service
> executor
> -------------------------------------------------------------------------------------
>
> Key: HDDS-9377
> URL: https://issues.apache.org/jira/browse/HDDS-9377
> Project: Apache Ozone
> Issue Type: Sub-task
> Components: OM, S3, SCM
> Affects Versions: 1.4.0
> Reporter: Vyacheslav Tutrinov
> Assignee: Alex Juncevich
> Priority: Major
> Labels: pull-request-available
> Attachments: Screenshot 2023-09-09 at 17.59.50.png
>
>
> We've made an experiment: load 15mlns of 1k files to the EC bucket and read
> them (3 clients in 50 threads => 150 threads in total). A high metrics
> related method contention was detected on
> org.apache.hadoop.metrics2.lib.MutableStat#add(long, long) (as it;s
> synchronized) - we hung for about a minute (sum) of a 10-minute profiling
> session
> !Screenshot 2023-09-09 at 17.59.50.png!
> The metrics-related operation results are not required at the time of
> execution, hence, they can be executed in another thread pool
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
