On 06/22/2009 01:25 PM, Alex Balashov wrote:
Daniel-Constantin Mierla wrote:
Hello,
On 06/22/2009 11:37 AM, Alex Balashov wrote:
[...]At the same time, the Kamailio leadership appears to have much more perspective and standardisation on rtpproxy, as well as whatever RTP forwarding module is being lifted from IPTel.org for SIP-Router.
rtpproxy does more now than packet forwarding (e.g., playing audio, moh, repacketization ...), therefore needs to stay in user space, see: http://www.rtpproxy.org.
I did not mean to suggest that it is a liability of rtpproxy that it is implemented in userspace.
Someone does need to do a study to see if kernel vs. userspace forwarding really makes that much of a difference, though. I suspect that although it does make a difference, it is not nearly as much of a difference as is commonly thought.
i think the same, although I haven't tested, rtpproxy is enough so far and can have many instances load balanced from kamailio nathelper module.
What I wanted to point out, is that nathelper is now more than rtp packages relayer and for that being in user space is necessary. When someone is looking at performances, iptrtpproxy module is another option using kernel space (whatever that makes any difference or not).
Cheers, Daniel
Perhaps I will do this study at Evariste and publish the results. The problem is I do not know of a good load-testing tool that generates bidirectional media. There is SIPP, but it is good for signaling and, at best, instantiating one-way media flows only, and also has the problem of making the far-end endpoint the limiting factor. Is it possible to set up SIPP or something like it (open source?) on two endpoints and make them set up calls between themselves with media in both directions?