> > Asterisk needs urgently to push the RTP engine to the Kernel, away from > userland, like professional and commercial softwares do. I measured the > cost of passing call from a public IP to a private IP, like typically a > Session Border Controller may do. In Asterisk, ulaw, no transcoding, it > takes 1.7% of a 3 Ghz core. If the packets where flowing through the > kernel, like iptables does, it would take 10% if the CPU. Asterisk then > could be used in hundreds of different roles in the enterprise. PJSIP has > no importance at all, this is the big issue. I suggest the developers look > at an open-source package and adapt the code, is called rtpengine. It uses > a kernel module to do the job.Philip -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.digium.com/pipermail/asterisk-users/attachments/20200512/0ed9473d/attachment.html>
patches welcome Philip Asterisk is Open Source so everyone can help Marek Dne 12/05/2020 v 07:02 Saint Michael napsal(a):> > Asterisk needs urgently to push the RTP engine to the Kernel, away > from userland, like professional and commercial softwares do. I > measured the cost of passing call from a public IP to a private > IP, like typically a Session Border Controller may do. In > Asterisk, ulaw, no transcoding, it takes 1.7% of a 3 Ghz core. If > the packets where flowing through the kernel, like iptables does, > it would take 10% if the CPU. Asterisk then could be used in > hundreds of different roles in the enterprise. PJSIP has no > importance at all, this is the big issue. I suggest the developers > look at an open-source package and adapt the code, is called > rtpengine. It uses a kernel module to do the job. > > Philip >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.digium.com/pipermail/asterisk-users/attachments/20200512/6075fa0d/attachment.html>