[asterisk-dev] [Code Review] 2873: Stasis performance improvements

[David Lee]

> On Sept. 20, 2013, 12:55 p.m., rmudgett wrote:
> > /branches/12/main/astobj2.c, lines 587-593
> > <https://reviewboard.asterisk.org/r/2873/diff/1/?file=46079#file46079line587>
> >
> >     Delete this as well.
> 
> David Lee wrote:
>     While I agree that the clause can probably be safely deleted, it wasn't pertinent to my changes.
> 
> rmudgett wrote:
>     This clause is the reason the memsets were writing as large as they were.  Since that code is no longer in the file, there is no reason for this code either.

All the same, I'm to scared to touch it :-D


- David


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/2873/#review9749
-----------------------------------------------------------


On Sept. 20, 2013, 4:05 p.m., David Lee wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/2873/
> -----------------------------------------------------------
> 
> (Updated Sept. 20, 2013, 4:05 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Repository: Asterisk
> 
> 
> Description
> -------
> 
> This patch addresses several performance problems that were found in
> the initial performance testing of Asterisk 12.
> 
> It turns out that we don't have many subscribers to Stasis topics in
> the system. We have lots of caching and forwarding, but those don't
> use the threadpool. Since we have such a limited number subscribers,
> it made sense to simply use taskprocessors and allocate a thread per
> subscriber instead.
> 
> The Stasis dispatch object was allocated as an AO2 object, even though
> it has a very confined lifecycle. This was replaced with a straight
> ast_malloc().
> 
> The Stasis message router was spending an inordinate amount of time
> searching hash tables. In this case, most of our routers had 6 or
> fewer routes in them to begin with. This was replaced with an array
> that's searched linearly for the route.
> 
> We more heavily rely on AO2 objects in Asterisk 12, and the memset()
> in ao2_ref() actually became noticeable on the profile. This was
> #ifdef'ed to only run when AO2_DEBUG was enabled.
> 
> Finally, since we're more reliant on the taskprocessor, it was
> optimized to reduce the number of times it was locking/unlocking the
> tps mutex.
> 
> 
> Diffs
> -----
> 
>   /branches/12/configs/stasis.conf.sample 399546 
>   /branches/12/include/asterisk/stasis.h 399546 
>   /branches/12/main/astobj2.c 399546 
>   /branches/12/main/stasis.c 399546 
>   /branches/12/main/stasis_config.c 399546 
>   /branches/12/main/stasis_message_router.c 399546 
>   /branches/12/main/taskprocessor.c 399546 
>   /branches/12/res/res_pjsip/include/res_pjsip_private.h 399546 
> 
> Diff: https://reviewboard.asterisk.org/r/2873/diff/
> 
> 
> Testing
> -------
> 
> Using the setup below, sipp was run against the 12 branch both with
> and without my changes.
> 
> ==
> [test-server]$ cat extensions.conf 
> 
> [performance]
> exten => service,1,Answer
> exten => service,n,Wait(1)
> exten => service,n,Hangup
> 
> [test-server]$ cat sip.conf 
> 
> [perf](!)
> type=peer
> qualify=no
> disallow=all
> allow=g722
> allow=ulaw
> insecure=invite,port
> context=performance
> 
> [test-client](perf)
> host=test-client
> 
> Simple sipp scenario:
> [test-client]$ sipp -sn uac -l 100 -r 50 tests-server
> 
> 
> Thanks,
> 
> David Lee
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20130920/416ac623/attachment-0001.htm>