[
https://issues.apache.org/jira/browse/HBASE-12988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14541369#comment-14541369
]
Lars Hofhansl commented on HBASE-12988:
---------------------------------------
Another potential way to speed this up (a little) is to build the next chunk of
edit to send while the current chunk is in flight. We'd double the required
heap on the replication source, but at least we can do some work on the source
while the current chunk is transmitted and applied at the source - but we'd
still only have exactly a single chunk outstanding at any given to guarantee
order.
And maybe, since we already have scenarios where edit for the same row can
arrive out of order for example when a region server fails, we can just make
the sink more robust and simply send edits out of order.
> [Replication]Parallel apply edits on row-level
> ----------------------------------------------
>
> Key: HBASE-12988
> URL: https://issues.apache.org/jira/browse/HBASE-12988
> Project: HBase
> Issue Type: Improvement
> Components: Replication
> Reporter: hongyu bi
> Assignee: hongyu bi
>
> we can apply edits to slave cluster in parallel on table-level to speed up
> replication .
> update : per conversation blow , it's better to apply edits on row-level in
> parallel
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
