[
https://issues.apache.org/jira/browse/CASSANDRA-2252?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13091229#comment-13091229
]
Yang Yang commented on CASSANDRA-2252:
--------------------------------------
actually even if we don't do the further optimization suggested in the last
comment (separate rowkey and column value into different slab allocators), it
would still very much likely work better and kill off some dead regions.
let's say a row/column is continually updated 1000 times , and 100 column
value fit into 2MB, then to do these 1000 updates, we allocate 10 regions, only
the first region would contain the row key, and finally all the 8 regions in
the middle would die, the first one remains due to the row key, and the last
remains due to the latest (live) column value
> arena allocation for memtables
> ------------------------------
>
> Key: CASSANDRA-2252
> URL: https://issues.apache.org/jira/browse/CASSANDRA-2252
> Project: Cassandra
> Issue Type: Improvement
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Fix For: 1.0
>
> Attachments: 0001-add-MemtableAllocator.txt,
> 0002-add-off-heap-MemtableAllocator-support.txt, 2252-v3.txt, 2252-v4.txt,
> merged-2252.tgz
>
>
> The memtable design practically actively fights Java's GC design. Todd
> Lipcon gave a good explanation over on HBASE-3455.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
