Delay and Packet Loss effects

Effect of delay and packet loss on VoIP when using FEC has been studied by many researchers [Jiang 2000], [Jiang 2002], [Jiang 2003], [Markopoulou 2003].

A rule of thumb: When the packet loss rate exceeds 20%, the audio quality of VoIP is degraded beyond usefulness (cited as [S03] in [Floyd 2004]).

Normally in telephony, when the quality falls below a certain level users give up (i.e., they hang up). Does this occur in the absence of a cost associated with not hanging up?

∴ according to [Floyd 2004]:

if loss rate is persistently unacceptably high relative to the current sending rate
& the best-effort application is unable to lower its sending rate:

⇒ flow must discontinue:

- multicast session ⇒ receiver withdraws from the multicast group
- unicast session ⇒ unicast connection termination

Slide Notes

Wenyu Jiang and Henning Schulzrinne, “Modeling of Packet Loss and Delay and Their Effect on Real-Time Multimedia Service Quality”, NOSSDAV, 2000. http://citeseer.nj.nec.com/jiang00modeling.html Links to an external site.

Wenyu Jiang and Henning Schulzrinne, “Comparison and Optimization of Packet Loss Repair Methods on VoIP Perceived Quality under Bursty Loss”, NOSSDAV, 2002. Available from http://www1.cs.columbia.edu/~wenyu/ Links to an external site.

Wenyu Jiang, Kazummi Koguchi, and Henning Schulzrinne, “QoS Evaluation of VoIP End-points”, ICC 2003. Available from http://www1.cs.columbia.edu/~wenyu/ Links to an external site.

A. P. Markopoulou, F. A. Tobagi, and M. J. Karam, “Assessing the Quality of Voice Communications Over Internet Backbones”, IEEE/ACM Transactions on Networking, V. 11 N. 5, October 2003, doi: 10.1109/TNET.2003.818179

S. Floyd and J. Kempf (Editors), “IAB Concerns Regarding Congestion Control for Voice Traffic in the Internet”, IETF, RFC 3714, Network Working Group, March 2004. ftp://ftp.rfc-editor.org/in-notes/rfc3714.txt Links to an external site.

Transcript

[slide473] The next thing was to say, well, we can use forward error correction so even when packets are dropped, we can still continue on. But, the rule of thumb is if the packet loss rate exceeds 20%, then it's probably degraded beyond useful. Now, previously we would have probably stopped the session when we got to 5% loss rate because the quality would have degraded. But with forward error correction, we could tolerate more lost packets. We still, at some point, need to back off. In normal telephoning, when the quality gets bad, what happens? The users give up and they try again later. What's the problem? These aren't humans, right? These are computers. So, Floyd in 2004 says if the loss rate is persistently unacceptably high, then we should make a best effort to lower it. And if we're still unable to lower our loss rate, the flow must be discontinued. So, we have to stop. Now, in multicast sessions, what happens? Well, the receiver withdraws from the multicast group. In a unicast session, we terminate the connection. We terminate our session. This all seems pretty straightforward, right? Everyone says, yeah, this makes sense. Hopefully. But what's the problem? When do we try again?