[jira] [Commented] (CASSANDRA-5966) Average name query performance much worse for wide rows
[ https://issues.apache.org/jira/browse/CASSANDRA-5966?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13759531#comment-13759531 ] Ravi Prasad commented on CASSANDRA-5966: This seems to be breaking HintedHandOffManager scheduled task, which does a empty names column filter: https://issues.apache.org/jira/browse/CASSANDRA-5979 > Average name query performance much worse for wide rows > --- > > Key: CASSANDRA-5966 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5966 > Project: Cassandra > Issue Type: Bug >Reporter: Rick Branson >Assignee: Rick Branson > Fix For: 1.2.10, 2.0.1 > > Attachments: 5966-trunk.txt, 5966.txt, 5966-v2.txt > > > The average performance for by-name queries degrades heavily on wide rows. > This is because in the wide row case SSTableNamesIterator deserializes every > column in the entire row chunk (64KB by default), where-as in the narrow row > case, it stops deserializing as soon as it's found the columns it's looking > for. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5966) Average name query performance much worse for wide rows
[ https://issues.apache.org/jira/browse/CASSANDRA-5966?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13755634#comment-13755634 ] Rick Branson commented on CASSANDRA-5966: - v2 Patch fixes the wide-row case where one of the query columns is an index entry's end range due to SortedSet.subSet() being end-exclusive. > Average name query performance much worse for wide rows > --- > > Key: CASSANDRA-5966 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5966 > Project: Cassandra > Issue Type: Bug >Reporter: Rick Branson >Assignee: Rick Branson > Attachments: 5966.txt, 5966-v2.txt > > > The average performance for by-name queries degrades heavily on wide rows. > This is because in the wide row case SSTableNamesIterator deserializes every > column in the entire row chunk (64KB by default), where-as in the narrow row > case, it stops deserializing as soon as it's found the columns it's looking > for. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
