You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@lucene.apache.org by "Feng Gui (Jira)" <ji...@apache.org> on 2021/08/02 04:05:00 UTC
[jira] [Updated] (LUCENE-10041) Duplication computation for
TieredMergePolicy's numDeletesToMerge
[ https://issues.apache.org/jira/browse/LUCENE-10041?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Feng Gui updated LUCENE-10041:
------------------------------
Description:
In lucene 8.4 (Elasticsearch 7.6.2), all indices are enabled with soft-deletes. We found the *TieredMergePolicy* contains redundant query computation to acquire *numDeletesToMerge* for *findMerges*. According to code from [getSortedBySegmentSize|https://github.com/apache/lucene/blob/main/lucene/core/src/java/org/apache/lucene/index/TieredMergePolicy.java#L303] , I think we should reuse the result of numDeletesToMerge which can reduce 3% CPU and some IO overheads.
.
was:
In lucene 8.4 (Elasticsearch 7.6.2), all indices are soft-deletes enabled. We found the *TieredMergePolicy* contains redundant query computation to acquire numDeletesToMerge for findMerges. According to code from [getSortedBySegmentSize|https://github.com/apache/lucene/blob/main/lucene/core/src/java/org/apache/lucene/index/TieredMergePolicy.java#L303] , we should reuse the result of numDeletesToMerge which can reduce 3% CPU and some IO overheads.
> Duplication computation for TieredMergePolicy's numDeletesToMerge
> -----------------------------------------------------------------
>
> Key: LUCENE-10041
> URL: https://issues.apache.org/jira/browse/LUCENE-10041
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/index
> Affects Versions: 8.4
> Reporter: Feng Gui
> Assignee: Nhat Nguyen
> Priority: Minor
> Attachments: IMG21.jpeg
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> In lucene 8.4 (Elasticsearch 7.6.2), all indices are enabled with soft-deletes. We found the *TieredMergePolicy* contains redundant query computation to acquire *numDeletesToMerge* for *findMerges*. According to code from [getSortedBySegmentSize|https://github.com/apache/lucene/blob/main/lucene/core/src/java/org/apache/lucene/index/TieredMergePolicy.java#L303] , I think we should reuse the result of numDeletesToMerge which can reduce 3% CPU and some IO overheads.
>
> .
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org
For additional commands, e-mail: issues-help@lucene.apache.org