[ https://issues.apache.org/jira/browse/CASSANDRA-6935?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13979001#comment-13979001 ]
Benedict commented on CASSANDRA-6935: ------------------------------------- This is another non-trivial patch as it will have to touch a lot of places, but the benefits are pretty substantial, and it's a necessary stepping stone on the way to bigger things. If we can figure out how to split it up into a few smaller pieces of work that would be great. > Make clustering part of primary key a first order component in the storage > engine > --------------------------------------------------------------------------------- > > Key: CASSANDRA-6935 > URL: https://issues.apache.org/jira/browse/CASSANDRA-6935 > Project: Cassandra > Issue Type: Improvement > Components: Core > Reporter: Benedict > Assignee: T Jake Luciani > Labels: performance > Fix For: 3.0 > > > It would be helpful for a number of upcoming improvements if the clustering > part of the primary key were extracted from CellName, and if a ColumnFamily > object could store multiple ClusteredRow (or similar) instances, within which > each cell is keyed only by the column identifier. > This would also, by itself, reduce on comparison costs and also permit memory > savings in memtables, by sharing the clustering part of the primary key > across all cells in the same row. It might also make it easier to move more > data off-heap, by constructing an off-heap clustered row, but keeping the > partition level object on-heap. -- This message was sent by Atlassian JIRA (v6.2#6252)