[ https://issues.apache.org/jira/browse/PHOENIX-1899?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Samarth Jain updated PHOENIX-1899: ---------------------------------- Attachment: PHOENIX-1899.patch Thinking about this more, this patch will end up creating scanners and calling next() on them even for cases when it was decided by phoenix that the query should run serially. Altered the patch to do this: {code} if (isSerial(context, table, orderBy, limit, allowPageFilter)) { return ParallelIteratorFactory.NOOP_FACTORY; } if ( ScanUtil.isRoundRobinPossible(orderBy, context)) { return ParallelIteratorFactory.FETCH_ON_CREATE_FACTORY; } {code} > Performance regression for non-aggregate, unordered queries returning 0 or > few records > -------------------------------------------------------------------------------------- > > Key: PHOENIX-1899 > URL: https://issues.apache.org/jira/browse/PHOENIX-1899 > Project: Phoenix > Issue Type: Bug > Reporter: Samarth Jain > Assignee: Samarth Jain > Attachments: PHOENIX-1899.patch > > > Details on here: > Apache Phoenix Performance > Suite: standard > Comparison: > v4.1.0,0.98.7-hadoop2;v4.2.2,0.98.7-hadoop2;v4.3.0,0.98.7-hadoop2;4.x-HBase-0.98,0.98.7-hadoop2 > Results: http://phoenix-bin.github.io/client/performance/latest.htm > History: http://phoenix-bin.github.io/client/performance -- This message was sent by Atlassian JIRA (v6.3.4#6332)