[
https://issues.apache.org/jira/browse/HBASE-12988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14590320#comment-14590320
]
Lars Hofhansl commented on HBASE-12988:
---------------------------------------
Any comments on this:
{code}
int n = Math.min(this.maxThreads, replicationSinkMgr.getSinks().size()/100+1);
{code}?
I'm trying to avoid breaking up very small batches. I.e. if the batch has only
50 edits to ship it does not make much sense to break that up into 5 chunks of
1 edit. Even when these edits are large... In that case the network time will
dominate and breaking up won't add much.
So I pulled 100 out of my hat.
Also maybe I should add a Math.min(..., number of sinks), so that we are not
using more threads than we have sinks region servers on the other side.
> [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: Lars Hofhansl
> Attachments: 12988.txt, HBASE-12988-0.98.patch,
> ParallelReplication-v2.txt
>
>
> 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)
