[asterisk-dev] [Code Review] 2890: Get rid of unnecessary waiting for endpoint and channel snapshots.

David Lee reviewboard at asterisk.org
Fri Sep 27 15:42:51 CDT 2013


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/2890/#review9845
-----------------------------------------------------------

Ship it!


Ship It!

- David Lee


On Sept. 27, 2013, 12:32 p.m., Mark Michelson wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/2890/
> -----------------------------------------------------------
> 
> (Updated Sept. 27, 2013, 12:32 p.m.)
> 
> 
> Review request for Asterisk Developers, David Lee and Joshua Colp.
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> A performance issue seen when doing a mid to high call-per-second test with chan_pjsip was that the device state callback was creating many stasis subscriptions, which in turn meant the creation of many taskprocessors, which in turn meant the creation of many threads.
> 
> Due to a change in when the stasis cache is updated (the publisher thread now updates the cache), the temporary subscriptions should no longer be required. The change presented here gets rid of the "guaranteed" parameter for ast_endpoint_latest_snapshot(). It also gets rid of the stasis_topic_wait() call in chan_pjsip.c since it should not be needed.
> 
> It may be possible to get rid of stasis_topic_wait() altogether, but I'm not so sure about that.
> 
> 
> Diffs
> -----
> 
>   /team/group/performance/channels/chan_pjsip.c 399935 
>   /team/group/performance/include/asterisk/stasis_endpoints.h 399935 
>   /team/group/performance/main/stasis_endpoints.c 399935 
>   /team/group/performance/res/ari/resource_endpoints.c 399935 
> 
> Diff: https://reviewboard.asterisk.org/r/2890/diff/
> 
> 
> Testing
> -------
> 
> Ran a 25 cps test that simply answered and hung up. These changes introduced about a 1% CPU performance gain on my system.
> 
> 
> Thanks,
> 
> Mark Michelson
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20130927/bf5de6a3/attachment-0001.html>


More information about the asterisk-dev mailing list