[jira] [Commented] (CASSANDRA-5389) Deserialize to arrays instead of trees
[ https://issues.apache.org/jira/browse/CASSANDRA-5389?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13630281#comment-13630281 ] Ahmed Bashir commented on CASSANDRA-5389: - How does this effect the overall memory footprint on the heap? Deserialize to arrays instead of trees -- Key: CASSANDRA-5389 URL: https://issues.apache.org/jira/browse/CASSANDRA-5389 Project: Cassandra Issue Type: Improvement Components: Core Reporter: Jonathan Ellis Assignee: Jonathan Ellis Priority: Minor Fix For: 2.0 Attachments: 5389.txt As near as I can tell, there's no reason to deserialize CFs to trees anymore. On the write path, we move it into an Atomic collection in the Memtable; on the read path, we also create a new clone when merging different results. -- 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-5389) Deserialize to arrays instead of trees
[ https://issues.apache.org/jira/browse/CASSANDRA-5389?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13630291#comment-13630291 ] Jonathan Ellis commented on CASSANDRA-5389: --- It's basically negligible since all of these are going to be swept out during young gen compaction, they'll never get tenured. The more important effect is that adding sorted entries to a TreeMap one at a time is the worst-case scenario for that Collection and can burn a lot of CPU rebalancing. Deserialize to arrays instead of trees -- Key: CASSANDRA-5389 URL: https://issues.apache.org/jira/browse/CASSANDRA-5389 Project: Cassandra Issue Type: Improvement Components: Core Reporter: Jonathan Ellis Assignee: Jonathan Ellis Priority: Minor Fix For: 2.0 Attachments: 5389.txt As near as I can tell, there's no reason to deserialize CFs to trees anymore. On the write path, we move it into an Atomic collection in the Memtable; on the read path, we also create a new clone when merging different results. -- 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-5389) Deserialize to arrays instead of trees
[ https://issues.apache.org/jira/browse/CASSANDRA-5389?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13615437#comment-13615437 ] Sylvain Lebresne commented on CASSANDRA-5389: - +1, as near as I can tell :). Tests are passing and I can't either find a place where we add columns out of order expect for memtables where we use an AtomicSortedColumns anyway. Deserialize to arrays instead of trees -- Key: CASSANDRA-5389 URL: https://issues.apache.org/jira/browse/CASSANDRA-5389 Project: Cassandra Issue Type: Improvement Components: Core Reporter: Jonathan Ellis Assignee: Jonathan Ellis Priority: Minor Fix For: 2.0 Attachments: 5389.txt As near as I can tell, there's no reason to deserialize CFs to trees anymore. On the write path, we move it into an Atomic collection in the Memtable; on the read path, we also create a new clone when merging different results. -- 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