[
https://issues.apache.org/jira/browse/IOTDB-5818?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17720837#comment-17720837
]
周沛辰 commented on IOTDB-5818:
----------------------------
h3. *Experiment*
----
Perform a retest experiment on the source files of this environment. The
sequential file size is 1.08G, the unsequential file size is 1.1 G, the
totalChunkNum in each file is 60200, each device has 301 alignment timeseries,
each timeseries has 2 chunks, and there are 100 devices in total. Execute cross
space compaction, actual occupied memory: < 900 M, forced frequent triggering
of Full GC, occupied memory < 300 M.
Originally: calculate Read UnseqFile occupies 56.32M, Read SeqFile occupies
1.5G, Write targetFile occupies 300M, total 1.8 G. Here the evaluation of Read
SeqFile is too large.
Improvement: Change the compression ratio to take from the memory, Read
UnseqFile occupies 37.54M, Read SeqFile occupies 18.432M, Write targetFile
occupies 300M, total 356M.
----
The following figure is a screenshot of memory usage during compaction.
!image-2023-05-09-15-40-32-907.png|width=1050,height=563!
> [Atmos][Compaction]Cross_space compaction of Aligned timeseries is stucked
> --------------------------------------------------------------------------
>
> Key: IOTDB-5818
> URL: https://issues.apache.org/jira/browse/IOTDB-5818
> Project: Apache IoTDB
> Issue Type: Bug
> Components: Core/Compaction
> Affects Versions: 1.1.0
> Reporter: Qingxin Feng
> Assignee: 周沛辰
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.1.1
>
> Attachments: image-2023-04-25-15-01-43-840.png,
> image-2023-05-09-15-40-32-907.png
>
>
> Cross_space compaction of Aligned timeseries is stucked since 0419 c97c0dc
> was merged
> !image-2023-04-25-15-01-43-840.png|width=646,height=214!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
