[
https://issues.apache.org/jira/browse/IOTDB-5263?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17650798#comment-17650798
]
Jinrui Zhang commented on IOTDB-5263:
-------------------------------------
We will do this optimization after bug fix in current implementation
> Optimize the cross compaction file selection and execution
> ----------------------------------------------------------
>
> Key: IOTDB-5263
> URL: https://issues.apache.org/jira/browse/IOTDB-5263
> Project: Apache IoTDB
> Issue Type: Improvement
> Reporter: Jinrui Zhang
> Assignee: Jinrui Zhang
> Priority: Major
> Attachments: image-2022-12-21-18-06-40-580.png
>
>
> In current implementation, when selecting the `overlapped` seq files for one
> specific unseq file, one seq file will always be selected even though it
> doesn't has overlap with the unseq file. See the sample below.
> !image-2022-12-21-18-06-40-580.png|width=718,height=220!
> That is, when selecting seq files for `3`, file-1 will be selected even
> though there is no overlap between 1 and 3. It is because we need to find a
> target file for 3 in current cross compaction implementation. Or there will
> be overlapped seq files generated after cross compaction.
> We need to do the optimization for it.
> # Only select the seq files which has overlapped with target unseq file.
> # change the implementation of the cross compaction to find the target seq
> file and decrease unnecessary file write.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
