The service class for the bearer stream, at least on modern configurations with Moto, is "DefVoiceDown" and "DefUGS". The signaling is DefRRDown and DefRRUp. MSOs may create different service classes with unique names, so our (plain vanilla configuration which uses the default names) may not be representative of other implementations. Frank -----Original Message----- From: Michael Thomas [mailto:mike@mtcc.com] Sent: Wednesday, March 02, 2011 10:36 AM To: Jay Ashworth Cc: NANOG Subject: Re: What vexes VoIP users? On 03/02/2011 06:23 AM, Jay Ashworth wrote:
----- Original Message -----
From: "Michael Thomas"<mike@mtcc.com>
Yes, really. The only difference was which L2 channels the RTP packets were flowed onto, which was determined by the MGCP/SIP signalling and interaction with the telephony gateway. There is a **very** complicated state machine that deals with this using some bastardized IETF protocols (COPS IIRC).
Ok, see, now I (like, I suspect, Frank Bulk) am confused again:
when you say "which L2 channels the RTP packets were flowed onto", that sounds to me a *whole* lot like "which VLAN on the end-user drop carried the RTP packets from the terminal adapter in their cable box to our concentrator"... which is pretty much the point I was originally trying to make, if perhaps in slightly different terms.
Am I still misunderstanding you?
They're kind of like VLAN's, but not exactly. It's been a long time since I worked on this... The RTP is flowed over what is called unsolicited grants (UGS) which give slots on the upstream for transmission. There are several other types of qos treatment between the CM and CMTS... I think that packetcable flows the MGCP and SIP over nrt-PS, but I might be misremembering. The signalling between the CM (MTA) and CMS (eg MGCP) is what fields the requests for better qos treatment for the RTP stream, and the CMS talks to the CMTS via COPS to set up the UGS flows to the cable modem/voice box (ie, an embedded MTA). In any case, this is 100% IP end to end, with all kinds of goodies for LEA and privacy to boot, which make the entire problem of faithfully reproducing the PSTN over IP a giant headache. Mike, I should know about the LEA aspect since I was the first one at Cisco to find and then dutifully step on that mine