[asterisk-dev] Asterisk 1.8 SIP_CAUSE performance regression

Paul Belanger pabelanger at digium.com
Mon Aug 22 09:56:33 CDT 2011


On 11-08-18 08:50 AM, Russell Bryant wrote:
> On Thu, Aug 18, 2011 at 8:42 AM, Matthew Nicholson
> <mnicholson at digium.com>  wrote:
>> Greetings,
>>
>> Recently a performance regression in chan_sip was discovered in Asterisk
>> 1.8. The regression is caused by chan_sip setting
>> MASTER_CHANNEL(HASH(SIP_CAUSE,<chan name>)) after each response received
>> on a channel. That feature has been made optional in the latest 1.8 SVN
>> code, but is currently still enabled by default. After some internal
>> discussion, we decided to consider disabling this feature by default in
>> future 1.8 versions. This would be an unexpected behavior change for
>> anyone depending on that SIP_CAUSE update in their dialplan.
>> Alternatively, with this feature enabled, anyone upgrading from Asterisk
>> 1.4 will see a 60% decrease in the amount of SIP traffic they can handle
>> before encountering problems.
>>
>> Before disabling this feature, we wanted to get a feel for how many
>> people are using it. If you use this feature, please respond to this
>> email and let us know.
>
> +1 on disabling it.  (with docs in UPGRADE.txt, release announcement)
>
If anybody else would like to comment on this issue, please do so. As 
thinks look now, I believe we'll be disabling this on asterisk 1.8 too.

Obviously, we'll need to make sure our release notes properly document this.

-- 
Paul Belanger
Digium, Inc. | Software Developer
twitter: pabelanger | IRC: pabelanger (Freenode)
Check us out at: http://digium.com & http://asterisk.org



More information about the asterisk-dev mailing list