[ https://issues.apache.org/jira/browse/IGNITE-8596?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16553088#comment-16553088 ]
ASF GitHub Bot commented on IGNITE-8596: ---------------------------------------- GitHub user AMashenkov opened a pull request: https://github.com/apache/ignite/pull/4413 IGNITE-8596: Avoid map requests to non-affinity index segments. You can merge this pull request into a Git repository by running: $ git pull https://github.com/gridgain/apache-ignite ignite-8596 Alternatively you can review and apply these changes as the patch at: https://github.com/apache/ignite/pull/4413.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #4413 ---- commit bbf64a298d4eeb7cae3b78c14d86302c530a73dc Author: Andrey V. Mashenkov <andrey.mashenkov@...> Date: 2018-07-23T16:14:51Z IGNITE-8596: Avoid map requests to non-affinity index segments. ---- > SQL: remove unnecessary index lookups when query parallelism is enabled > ----------------------------------------------------------------------- > > Key: IGNITE-8596 > URL: https://issues.apache.org/jira/browse/IGNITE-8596 > Project: Ignite > Issue Type: Task > Components: sql > Affects Versions: 2.5 > Reporter: Vladimir Ozerov > Assignee: Andrew Mashenkov > Priority: Major > Labels: performance > Fix For: 2.7 > > > See > {{org.apache.ignite.internal.processors.query.h2.twostep.GridMapQueryExecutor#onQueryRequest}} > method. If table is segmented, we will submit as many SQL requests as much > segments. But consider a case when target cache partition(s) is already > defined by user or derived through partition pruning. In this case most of > segments will not contain useful information and return empty result set. At > the same time these queries may impose index or data page scans, thus > consuming resources without a reason. > To mitigate the problem we should not submit SQL requests to segments we are > not interested in. > Note that it is not sufficient to simply skip SQL requests on mapper, because > reducer expects separate response for every message. We should fix both local > mapper logic as well as protocol. -- This message was sent by Atlassian JIRA (v7.6.3#76005)