[ https://issues.apache.org/jira/browse/CASSANDRA-3617?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13270147#comment-13270147 ]
Brandon Williams commented on CASSANDRA-3617: --------------------------------------------- In a 4 node cluster with rf=1 (to maximize the number of messages needed) using 4M rows there was about a 6% improvement on writes and a 12.5% improvement on reads. I expect this would grow over time with the size of the test. > Clean up and optimize Message > ----------------------------- > > Key: CASSANDRA-3617 > URL: https://issues.apache.org/jira/browse/CASSANDRA-3617 > Project: Cassandra > Issue Type: Improvement > Components: Core > Reporter: Jonathan Ellis > Assignee: Yuki Morishita > Fix For: 1.2 > > > The Message class has grown largely by accretion and it shows. There are > several problems: > - Outbound and inbound messages aren't really the same thing and should not > be conflated > - We pre-serialize message bodies to byte[], then copy those bytes onto the > Socket buffer, instead of just keeping a reference to the object being > serialized and then writing it out directly to the socket > - MessagingService versioning is poorly encapsulating, scattering version > variables and references to things like CachingMessageProducer across the > codebase -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira