[
https://issues.apache.org/jira/browse/HDDS-9784?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17792102#comment-17792102
]
Hongbing Wang commented on HDDS-9784:
-------------------------------------
[~XiChen] I also tested the result of different parameters in a 4 datanodes
cluster.
{code:java}
ozone freon ommg --operation CREATE_KEY -n 10000000 --volume=s3v --bucket=test1
-t 50 --size 4096
{code}
{code:java}
(1) hdds.ratis.raft.server.log.appender.wait-time.min=0
4.89% |██ | 488892/10000000 Time: 0:02:36|
CREATE_KEY: rate 4642 max 4681
(2) hdds.ratis.raft.server.log.appender.wait-time.min=1us
4.31% |██ | 430701/10000000 Time: 0:02:19|
CREATE_KEY: rate 3633 max 4227
(3) hdds.ratis.raft.server.log.appender.wait-time.min=5us
4.24% |██ | 423695/10000000 Time: 0:02:07|
CREATE_KEY: rate 3799 max 4756
(4) hdds.ratis.raft.server.log.appender.wait-time.min=1ms
3.17% |██ | 316660/10000000 Time: 0:02:44|
CREATE_KEY: rate 2236 max 2446
{code}
1ms performance is much worse, 0 and 1us, 5us is not much different. Perhaps
the empirical values you have chosen are reasonable.
> Datanode Write performance degradation
> --------------------------------------
>
> Key: HDDS-9784
> URL: https://issues.apache.org/jira/browse/HDDS-9784
> Project: Apache Ozone
> Issue Type: Bug
> Components: Ozone Datanode
> Reporter: ChenXi
> Priority: Blocker
> Labels: pull-request-available
> Attachments: image-2023-11-28-15-12-49-200.png,
> image-2023-11-28-15-13-39-193.png, image-2023-11-28-15-14-12-518.png,
> image-2023-11-28-15-17-47-458.png, image-2023-11-28-15-18-12-206.png,
> image-2023-11-28-15-21-49-301.png, image-2023-11-29-11-00-42-782.png,
> image-2023-11-29-11-05-49-209.png, image-2023-11-29-11-06-30-963.png,
> image-2023-11-29-11-07-01-321.png, image-2023-11-29-13-19-18-869.png,
> image-2023-11-29-19-31-07-482.png, image-2023-11-29-19-45-03-787.png,
> image-2023-11-29-19-45-10-996.png, image-2023-11-29-19-45-31-747.png,
> image-2023-11-29-19-47-18-429.png
>
>
> h2. Summary
> Default value of hdds.ratis.raft.server.log.appender.wait-time.min 1ms causes
> write new energy to degrade.
>
> h2. Testing commands
> Datanode Write performance degradation:
> command
> ozone freon ommg --operation CREATE_KEY -n 10000000 --volume=s3v
> --bucket=bucket-object -t 50 --size 4k
> h2. Bad Performance
> <property>
> <name>hdds.ratis.raft.server.log.appender.wait-time.min</name>
> <value>1ms</value>
> </property>
> !image-2023-11-28-15-14-12-518.png!
>
> h2. Normal Performance
> <property>
> <name>hdds.ratis.raft.server.log.appender.wait-time.min</name>
> <value>0</value>
> </property>
> !image-2023-11-28-15-13-39-193.png!
> !image-2023-11-29-13-19-18-869.png!
> h2. Other testing data can be referred to.
> 0 size performance, about 8000 QPS, this represents the highest write
> performance of any OM.
> the `hdds.ratis.raft.server.log.appender.wait-time.min` of om had been set to
> 0ms by default.
> !image-2023-11-28-15-18-12-206.png!
>
> This represents the performance limit of the DN, about 6000 QPS, which can
> indicate that the cluster itself has normal read/write performance.
> !image-2023-11-28-15-21-49-301.png!
>
> h2. Cluster info
> * Number of Datanodes 6
> * 12disks per Datanode
> * HA mode for both OM and SCM
> * Ozone version: ozone.git -r 60bb0604217fbd44dbc6b115b59fc26a297fcaba
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
