Alan, This is a good description of the requirements. The question I had was how to convey it to the mid-flow entity who do not have access to the RTCP. Is it by IRR which is H.323 specific or do we need something that will work both for H.323 and SIP. Roni
-----Original Message----- From: Alan Clark [mailto:alan@telchemy.com] Sent: Thursday, December 27, 2001 5:42 PM To: paulej@packetizer.com; Roni Even (E-mail) Subject: FW: Including RTCP report packets in IRR
Paul/ Roni
I tried posting this to the SG16 reflector but it didn't get sent out .......
Alan
-----Original Message----- From: Alan Clark [mailto:alan@telchemy.com] Sent: Wednesday, December 26, 2001 9:36 AM To: 'ITU-SG16@echo.jf.INTEL.COM' Subject: RE: Including RTCP report packets in IRR
Paul/ Roni/ Frank
We have been introducing this general topic via ETSI TIPHON and more recently presented at the IETF-52 AVT session. We believe that RTCP stats are not sufficient for VoIP purposes and also share the concern expressed by others on this thread that it is vital to have several approaches to getting data from the endpoints.
The approach we advocate is:-
(i) Calculate call quality metrics in the endpoint It is necessary to consider the "burstiness" of packet loss in order to assess impact on VoIP call quality - PLC is effective at hiding isolated lost packets but not bursts. VQmon (ETSI TS101 329-5 Annex E) is a lightweight E model based algorithm for estimating call quality that incorporates a burst packet loss model.
(ii) Recognize that there are several "audiences" for the data In some networks both the network operations/ management and billing groups would like to have access to call quality data however their requirements are different.
- operations/management .. real time event reporting, SNMP access, test and diagnosis .. may need access to data during a call .. needs support for fault isolation
- billing .. end of call metrics .. feed back through end of call messages
(iii) Provide several ways to get data out We would like to see multiple ways of extracting data both from the end-points and mid-flow. These should include extended RTCP reports with call quality metrics, end of call (DRQ or equivalent) and SNMP access.
Alan Clark Telchemy
-----Original Message----- From: Mailing list for parties associated with ITU-T Study Group 16 [mailto:ITU-SG16@echo.jf.INTEL.COM]On Behalf Of Even ,Roni Sent: Tuesday, December 25, 2001 9:33 AM To: ITU-SG16@echo.jf.INTEL.COM Subject: Re: Including RTCP report packets in IRR
Paul, Having a solution for H.323 may not solve the problem. RTP & RTCP are used also in other protocols like SIP. At the last IETF meeting (IETF-52) the topic was discussed in the AVT WG. Maybe we can see what will happen there. I would like to point out that SNMP can be used to get part of the information. Roni Even
-----Original Message----- From: Paul E. Jones [mailto:paulej@PACKETIZER.COM] Sent: Monday, December 24, 2001 3:14 AM To: ITU-SG16@echo.jf.INTEL.COM Subject: Re: Including RTCP report packets in IRR
Paul,
I suppose this is probably in need of clarification, but the assumption was that if the call contained an audio session, the audio element would be included. Likewise for video, etc. Unfortunately, there are a few fields in H.225.0 that are labeled as "OPTIONAL" in the ASN.1, so implementers assume that semantically they're optional, but that's not the case. We don't want to change the syntax (since we don't want to mandate the inclusion of a field when the call doesn't have a session of that type), so I suppose it would be nice to see a proposal for clarification of this issue.
One other thought about the RTCP stats is that we report them in the DRQ rather than the IRR. One problem with IRRs is that if they are sent too frequently, an extremely busy system can get bogged down processing those messages. I imagine that you'd want to set the IRR timer somewhere outside the length of an average call. Doing so, though, would mean you never get stats for the average call. So perhaps DRQ alone (or in addition to) the IRR is the right approach.
Paul
----- Original Message ----- From: "Paul Long" plong@IPDIALOG.COM To: ITU-SG16@echo.jf.INTEL.COM Sent: Tuesday, December 18, 2001 10:42 AM Subject: Re: Including RTCP report packets in IRR
Paul,
Please note though that an EP is not required to include any media info in perCallInfo, i.e., audio, video, and data. I imagine that this was an oversight, but we ought to explicitly require v5+ EPs to report all currently active media sessions, including RTP.
Paul Long ipDialog, Inc.
-----Original Message----- From: Mailing list for parties associated with ITU-T Study Group 16 [mailto:ITU-SG16@echo.jf.INTEL.COM]On Behalf Of Paul E. Jones Sent: Tuesday, December 18, 2001 3:07 AM To: ITU-SG16@echo.jf.INTEL.COM Subject: Re: Including RTCP report packets in IRR
I completely support your idea.
pj
----- Original Message ----- From: frank.derks@PHILIPS.COM To: ITU-SG16@echo.jf.INTEL.COM Sent: Monday, December 17, 2001 2:42 AM Subject: Including RTCP report packets in IRR
Dear all,
RTP's specification mandates that the RTP ports, for a given source/destination, are "co-located" with the RTCP ports. This means the the report information about the RTP streams flow between the communicating endpoints. Although this mode of operation is intentional, it does make it difficult for 3rd parties to monitor the statistics as perceived by the endpoints. This can be particularly useful for network management where one would like to be able to monitor whether the network provides adequate quality.
On way to access the RTCP information would be to route all the RTCP traffic through some entity, but this also means that the RTP will have to follow the same path through the network. And this is not desirable.
For H.323 there would be the option to include the report packets in the "audio" element in the perCallInfo element in the IRR message. This would allow for a Gatekeeper to get a hold of these "vital" statistics.
Any views?
Regards,
Frank