[
https://issues.apache.org/jira/browse/HBASE-9868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13810455#comment-13810455
]
Nicolas Liochon commented on HBASE-9868:
----------------------------------------
bq. GCing
I've seen it as well in my test. I'm profiling the client w/ 3 servers. I had
no OOME, but it seems we're creating a lot of objects. Yourkit was telling me
that the threads were blocked (so it's not only time or memory) in
getCachedLocation.
I wonder if we don't have more object since we protobuffed the client. We spend
a lot of time there. Our MultiAction adds little value: if we want it to be
efficient, we need to have the same structure as the protobuf message anyway so
we don't abstract much.
> Remove some array copy, especially around protobuf
> --------------------------------------------------
>
> Key: HBASE-9868
> URL: https://issues.apache.org/jira/browse/HBASE-9868
> Project: HBase
> Issue Type: Bug
> Components: Client, Protobufs
> Affects Versions: 0.98.0, 0.96.0
> Reporter: Nicolas Liochon
> Assignee: Nicolas Liochon
> Fix For: 0.98.0, 0.96.1
>
> Attachments: 9868.v1.patch
>
>
> Profiling the client shows that we're spending some time in array copy (10%
> of the code execution, 3% of the total time) in some array copy that we can
> avoid.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
