webrtc-stats packetsDiscarded and "effective packet loss"

833 views
Skip to first unread message

chris...@sinch.com

unread,
Oct 24, 2016, 5:48:42 AM10/24/16
to discuss-webrtc
I'm interested in being able to query for the stats metric |packetsDiscarded|, defined in https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-packetsdiscarded as "The cumulative number of RTP packets discarded by the jitter buffer due to late or early-arrival, i.e., these packets are not played out".

Is there any ETA for when |packetsDiscarded| will be implemented?


Actually, what I'm really after is a metric that can represent "effective packet loss", i.e. not only packet loss on network transmission level (e.g. what is covered by the metric |packetsLost|), but loss due to both network + late arrival (and thus discards in jitter buffer / NetEq). Ideally I would like to depend only on stats that are defined in the WebRTC spec (draft, but still), i.e. rather than Google-specific 'goog'-prefixed stats. But, I would also be interested in an opinion on whether it's plausible and reasonable to e.g. use the NetEq PLC stats as proxy metrics for "effective packet loss". E.g. I'm thinking of using the Google-specific "googDecoding-" metrics, and more specifically something like the ratio  (googDecodingPLC + googDecodingPLCCNG) / googDecodingCTN  as a proxy metric for effective packet loss. 


Thanks / C

Christoffer Jansson

unread,
Oct 25, 2016, 10:57:21 AM10/25/16
to discuss-webrtc
Hi,

You can file WebRTC stats spec issues/requests and/or questions here https://github.com/w3c/webrtc-stats.

/Chris

--

---
You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrt...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/discuss-webrtc/ee69eb6d-91e8-4eed-b8c9-f9a330f186b4%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
--
/Chris

Christoffer Jansson

unread,
Oct 25, 2016, 10:57:50 AM10/25/16
to discuss-webrtc, h...@google.com
+Harald Alvestrand do you know when packetsDiscarded will be implemented?
--
/Chris

Harald Alvestrand

unread,
Oct 25, 2016, 11:01:46 AM10/25/16
to Christoffer Jansson, Henrik Boström, discuss-webrtc
+hbos may have a clearer idea.



On Tue, Oct 25, 2016 at 4:57 PM, Christoffer Jansson <jan...@google.com> wrote:
+Harald Alvestrand do you know when packetsDiscarded will be implemented?
On Tue, Oct 25, 2016 at 4:56 PM Christoffer Jansson <jan...@google.com> wrote:
Hi,

You can file WebRTC stats spec issues/requests and/or questions here https://github.com/w3c/webrtc-stats.

/Chris

On Mon, Oct 24, 2016 at 11:48 AM <chris...@sinch.com> wrote:
I'm interested in being able to query for the stats metric |packetsDiscarded|, defined in https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-packetsdiscarded as "The cumulative number of RTP packets discarded by the jitter buffer due to late or early-arrival, i.e., these packets are not played out".

Is there any ETA for when |packetsDiscarded| will be implemented?


Actually, what I'm really after is a metric that can represent "effective packet loss", i.e. not only packet loss on network transmission level (e.g. what is covered by the metric |packetsLost|), but loss due to both network + late arrival (and thus discards in jitter buffer / NetEq). Ideally I would like to depend only on stats that are defined in the WebRTC spec (draft, but still), i.e. rather than Google-specific 'goog'-prefixed stats. But, I would also be interested in an opinion on whether it's plausible and reasonable to e.g. use the NetEq PLC stats as proxy metrics for "effective packet loss". E.g. I'm thinking of using the Google-specific "googDecoding-" metrics, and more specifically something like the ratio  (googDecodingPLC + googDecodingPLCCNG) / googDecodingCTN  as a proxy metric for effective packet loss. 


Thanks / C

--

---
You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrtc+unsubscribe@googlegroups.com.
--
/Chris
--
/Chris

Senthil Ganesh S

unread,
Aug 14, 2018, 9:33:23 AM8/14/18
to discuss-webrtc
When can we expect this?


On Tuesday, 25 October 2016 20:31:46 UTC+5:30, Harald Alvestrand wrote:
+hbos may have a clearer idea.


On Tue, Oct 25, 2016 at 4:57 PM, Christoffer Jansson <jan...@google.com> wrote:
+Harald Alvestrand do you know when packetsDiscarded will be implemented?
On Tue, Oct 25, 2016 at 4:56 PM Christoffer Jansson <jan...@google.com> wrote:
Hi,

You can file WebRTC stats spec issues/requests and/or questions here https://github.com/w3c/webrtc-stats.

/Chris

On Mon, Oct 24, 2016 at 11:48 AM <chris...@sinch.com> wrote:
I'm interested in being able to query for the stats metric |packetsDiscarded|, defined in https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-packetsdiscarded as "The cumulative number of RTP packets discarded by the jitter buffer due to late or early-arrival, i.e., these packets are not played out".

Is there any ETA for when |packetsDiscarded| will be implemented?


Actually, what I'm really after is a metric that can represent "effective packet loss", i.e. not only packet loss on network transmission level (e.g. what is covered by the metric |packetsLost|), but loss due to both network + late arrival (and thus discards in jitter buffer / NetEq). Ideally I would like to depend only on stats that are defined in the WebRTC spec (draft, but still), i.e. rather than Google-specific 'goog'-prefixed stats. But, I would also be interested in an opinion on whether it's plausible and reasonable to e.g. use the NetEq PLC stats as proxy metrics for "effective packet loss". E.g. I'm thinking of using the Google-specific "googDecoding-" metrics, and more specifically something like the ratio  (googDecodingPLC + googDecodingPLCCNG) / googDecodingCTN  as a proxy metric for effective packet loss. 


Thanks / C

--

---
You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrt...@googlegroups.com.
--
/Chris
--
/Chris

Reply all
Reply to author
Forward
0 new messages