You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@lucene.apache.org by "Cassandra Targett (Jira)" <ji...@apache.org> on 2020/02/28 17:13:00 UTC
[jira] [Commented] (SOLR-12407) edismax boost performance
regression from switch to FunctionScoreQuery
[ https://issues.apache.org/jira/browse/SOLR-12407?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17047805#comment-17047805 ]
Cassandra Targett commented on SOLR-12407:
------------------------------------------
[~romseygeek] or anyone else following this - can anyone confirm if this was resolved by SOLR-13126 as the last comment from a year ago suggests?
> edismax boost performance regression from switch to FunctionScoreQuery
> ----------------------------------------------------------------------
>
> Key: SOLR-12407
> URL: https://issues.apache.org/jira/browse/SOLR-12407
> Project: Solr
> Issue Type: Bug
> Affects Versions: 7.3
> Reporter: Will Currie
> Priority: Minor
> Attachments: restore-boosted-query.patch, solr-7.2.svg, solr-7.3.svg
>
>
> Assertion: FunctionScoreQuery uses the iterator style API (advanceExact + doubleValue). BoostedQuery uses the "old" api (just a single call to doubleValue). In an edismax boost this means the boost function is called twice for every document being scored in 7.3 instead of once in 7.2.
> I'm seeing ~50% increase in query response time after upgrading from 7.2 to 7.3 (600ms to 900ms). My queries use an edismax boost something like:
> {noformat}
> if(termfreq(type,"A"),product(map(field1,3,3,1.5,1),map(field1,4,4,1.9,1),if(def(field2,false),product(map(field1,1,1,0.6,1),map(field1,2,2,0.7,1),if(not(exists(field1)),0.6,1),map(field3,0,0,1.3,1)),product(map(field1,1,1,0.7,1),map(field1,2,2,1.1,1),if(not(exists(field1)),0.90,1),map(field3,0,0,1.50,1)))),1){noformat}
> This boost is likely (surely?) suboptimal but LUCENE-8099 appears to have introduced this performance regression (poured proverbial oil on my smouldering fire). If I change ExtendedDismaxQParser back to using the deprecated BoostedQuery I get the 600ms solr 7.2 response time back.
> It appears FunctionScoreQuery invokes the boost function twice for each document. Once with a call to [exists()|https://github.com/apache/lucene-solr/blob/03afeb7766a39996de3c85e8a6ab24d2a352dd1c/lucene/queries/src/java/org/apache/lucene/queries/function/ValueSource.java#L150] from [advanceExact()|https://github.com/apache/lucene-solr/blob/42154387d4f2a6060da09c4236e2a8dbb575c59e/lucene/queries/src/java/org/apache/lucene/queries/function/FunctionScoreQuery.java#L170], then a second time from the call chain following scores.doubleValue().
> I don't know if that's the cause of the slowdown but I'm definitely seeing a slowdown that disappears when I revert part of LUCENE-8099.
> I've attached some flamegraphs comparing 7.2 and 7.3. The frame FunctionScoreQuery$FunctionScoreWeight$1.score in solr-7.3.svg show 2 "towers". One for advanceExact (calling exists()), the other for doubleValue() which ends up similar to solr-7.2.svg.
--
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