[
https://issues.apache.org/jira/browse/POOL-277?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14138863#comment-14138863
]
Bernd Eckenfels commented on POOL-277:
--------------------------------------
I have already commented on the PR (at least getMean() should be getMax()). But
I wonder if the complicated queueing can be avoided with a lockless max
alternative using AtomicLong like:
{code}
do {
currentMax = maxWaitTime.get();
if (waitTime < currentMax)
break;
} while (!maxWaitTime.compareAndSet(currentMax, waitTime))
{code}
This asumes the if typically breaks the loop early (no new max) and a new max
can have some more actions as the system is slow in that case anyway. This
totally avoids the additional class and risky list maintenance.
> Monitor contention at BaseGenericObjectPool
> -------------------------------------------
>
> Key: POOL-277
> URL: https://issues.apache.org/jira/browse/POOL-277
> Project: Commons Pool
> Issue Type: Improvement
> Affects Versions: 2.2
> Reporter: Lucas Pouzac
> Labels: concurrency, performance
> Fix For: 2.3
>
> Attachments: monitor_contention_basegenericobjectpool.png,
> monitor_contention_basegenericobjectpool_after_pull_request.png
>
>
> Updating JMX causes synchronization while the service is secondary. See
> screenshot.
> I create a Pull Request [#1|https://github.com/apache/commons-pool/pull/1].
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
