[jira] [Updated] (SLING-4999) Check synchronization in RequestProcessorMBeanImpl

["Robert Munteanu (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/SLING-4999?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Robert Munteanu updated SLING-4999:
-----------------------------------
    Attachment: 0001-SLING-4999-Check-synchronization-in-RequestProcessor.patch

I've attached  patch which removes all synchronisation and instead stores data in an immutable object help by an atomic reference.

There should be no locking anymore, at the expense of (slightly) more computation done per request.

Due to the small response time I was unable to get meaningful performance results with my test run.

[~cziegeler], thought? Also [~joelrich] might be interested, he's doing lots of performance work recently

> Check synchronization in RequestProcessorMBeanImpl
> --------------------------------------------------
>
>                 Key: SLING-4999
>                 URL: https://issues.apache.org/jira/browse/SLING-4999
>             Project: Sling
>          Issue Type: Improvement
>          Components: Engine
>    Affects Versions: Engine 2.4.2
>            Reporter: Carsten Ziegeler
>             Fix For: Engine 2.4.4
>
>         Attachments: 0001-SLING-4999-Check-synchronization-in-RequestProcessor.patch
>
>
> Each and every request goes through RequestProcessorMBeanImpl#addRequestData which is a synchronized block. I think this is a rather heavy bottleneck for request processing.
> In addition, some of the methods which do additional calculation are sync'ed (like getStandardDeviationDurationMsec) but others are not (like getStandardDeviationPeakRecursionDepth). This seems a little bit arbitrary to me



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)