1281035 - Firefox WebRTC - High CPU usage in device (microphone/camera) renegotiation

Status: UNCONFIRMED

(Core :: WebRTC: Networking, defect, P3)

Description

[Abhinav Khawarey]

Attachment: FF47_log.txt.zip

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_11_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/52.0.2743.41 Safari/537.36

Steps to reproduce:

1. Join session by providing Guest Name on https://apollodemo01.bbcollab.com/collab/ui/session/guest/8A5D356DFE1118DC9196240915B8CD23?redialOnFFTC=false
2. Start microphone.
3. Check CPU Usage on Activity Monitor for Firefox.

Attaching signaling logs - FF47_log.txt.zip, 
about:webrtc page output - about-webrtc.txt,
Screenshot - Firefox 47 Process High CPU WebRTC.png
Macbook details - MacBook Pro Abhinav.zip


Actual results:

CPU usage for Firefox process spikes to ~500% after starting microphone.


Expected results:

CPU usage should never spike to ~500%.

Comment 1

[Abhinav Khawarey]

Attachment: about-webrtc.txt

Comment 2

[Abhinav Khawarey]

Attachment: Firefox 47 Process High CPU WebRTC .png

Comment 3

[Abhinav Khawarey]

Attachment: MacBook Pro Abhinav.zip

Comment 4

[Abhinav Khawarey]

Uploading trace and accompanying signaling logs - Instruments FF47.trace.zip and FF47_log_2.txt.zip

Comment 5

[Abhinav Khawarey]

Attachment: FF47_log_2.txt.zip

Comment 6

[Abhinav Khawarey]

Instruments FF47.trace.zip available at https://drive.google.com/file/d/0BzQI6MLbgP7vdUI1YnFtS2dRSWc/view for download.

Comment 7

[Balwant Bisht]

We see this issue only when user accesses devices and SDP is renegotiated. If we re-establish the PeerConnection in place of renegotiation we don't see high CPU consumption.

Comment 8

[Emma Humphries ☕️🎸🧞‍♀️✨ (she/they) [:emceeaich] (Pacific Time) use needinfo]

This is a P1 bug without an assignee. 

P1 are bugs which are being worked on for the current release cycle/iteration/sprint. 

If the bug is not assigned by Monday, 28 August, the bug's priority will be reset to '--'.

Comment 9

[Bulk Bug Changes for mreavy's org]

Mass change P1->P2 to align with new Mozilla triage process

Comment 10

[Sylvestre Ledru [:Sylvestre]]

Moving to p3 because no activity for at least 1 year(s).
See https://github.com/mozilla/bug-handling/blob/master/policy/triage-bugzilla.md#how-do-you-triage for more information

Comment 11

[Frederick Zhang]

Bump. This drains the battery of my laptop every time I make a conference call (which I do relatively often). It'd be nice if this could be fixed.

Comment 12

[Byron Campen [:bwc]]

I cannot reproduce the problem in the description; CPU usage does not look unusual.

frederick888, which conferencing service are you using? Is this an audio-only conference, or audio/video? How many participants?

Comment 13

[Frederick Zhang]

(In reply to Byron Campen [:bwc] from comment #12)

I cannot reproduce the problem in the description; CPU usage does not look unusual.

frederick888, which conferencing service are you using? Is this an audio-only conference, or audio/video? How many participants?

I experienced this issue when using UberConference and Google Hangouts Meet. They were audio-only calls with 3-5 people.

By the way I'm using an old Macbook (model A1398 EMC 2881) with integrated Intel graphics and Nvidia.

Comment 14

[Frederick Zhang]

Attachment: cpu usage linux.png

Comment 15

[Frederick Zhang]

Attachment: cpu usage macos.png

Comment 16

[Frederick Zhang]

I made a UberConf call between my Macbook and Linux laptop and it seems that it's even worse in Linux (see in attached screenshots).

My Linux laptop is using an Intel i7-6820HK and the integrated graphics is disabled.

Comment 17

[Byron Campen [:bwc]]

Ok, so that does not really look like what was in the description. It could be more efficient though. It is possible that the conferencing service you're using work with a full-mesh, meaning you have to encode the same audio separately for each participant, which can get quite expensive. I would be able to tell from the output of about:webrtc.