[
https://issues.apache.org/jira/browse/HADOOP-10278?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13890056#comment-13890056
]
Chris Li commented on HADOOP-10278:
-----------------------------------
[~daryn]
Implemented and benchmarked it. All around less clutter and custom locking
required. I do like this solution.
Benchmarked performance:
using RPCCallBenchmark -s 4 -c 10
Average server time increased by 1.5% (0.05 p value,
1-tailed/2-sample-equal-variance t test)
Other results were insignificant
If the performance hit is acceptable, I think this solution is easier, and I
can re-submit this patch.
-----
Raw results
Group calls/s Server CPU Client CPU
Poll 32356.0 48018 47837
Poll 31497.0 46129 44052
Poll 32456.0 46400 43643
Poll 31253.0 46343 44152
Poll 31572.0 46241 44123
Take 30831.0 46521 45315
Take 31764.0 45549 43827
Take 32534.0 45949 44028
Take 32153.0 45699 44215
Take 31455.0 45823 43926
T Tests:
Throughput 0.419851632864817
Server CPU 0.0509472764018302
Client CPU 0.279864716054668
> Refactor to make CallQueue pluggable
> ------------------------------------
>
> Key: HADOOP-10278
> URL: https://issues.apache.org/jira/browse/HADOOP-10278
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: ipc
> Reporter: Chris Li
> Attachments: HADOOP-10278.patch, subtask1.3.patch, subtask1.4.patch,
> subtask1.5.patch
>
>
> * Refactor CallQueue into an interface, base, and default implementation that
> matches today's behavior
> * Make the call queue impl configurable, keyed on port so that we minimize
> coupling
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
