Create experiment for sparse BWE updates

When receiving an RTCP packet containing feedback about multiple RTP packets,
we currently check for bandwidth overuse once for every RTP packet.

This CL creates an experiment to test processing all packets in the RTCP
feedback before checking for overuse. This can be thought of as checking
for overuse per RTCP packet instead of per RTP packet.

The change is not expected to have a large impact, but enabling the
experiment will make the delay-based BWE slightly less sensitive. This means
that we'll be less likely to back down incorrectly after a brief network
transient, at the cost of sometimes missing real overuse (especially when
the network queues are short). In the latter case, the loss-based estimator
is expected to detect the overuse.

The experiment is off by default.

BUG=webrtc:7508
Review-Url: https://codereview.webrtc.org/2835573003
Cr-Commit-Position: refs/heads/master@{#17968}
Committed: https://chromium.googlesource.com/external/webrtc/+/bf2c049a12302141725af0a7407022908fda7efb

[terelius, 2017-04-21 10:02:02]

The CQ bit was checked by terelius@webrtc.org to run a CQ dry run

[commit-bot: I haz the power, 2017-04-21 10:02:07]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.webrtc.org/...

[terelius, 2017-04-21 10:03:09]

Description was changed from

==========
When receiving an RTCP packet containing feedback about multiple RTP packets,
we currently check for bandwidth overuse once for every RTP packet.

This CL creates an experiment to test processing all packets in the RTCP
feedback before checking for overuse. This can be thought of as checking
for overuse per RTCP packet instead of per RTP packet.

The change is not expected to have a large impact, but enabling the
experiment will make the delay-based BWE slightly less sensitive. This means
that we'll be less likely to back down incorrectly after a brief network
transient, at the cost of sometimes missing real overuse (especially when
the network queues are short). In the latter case, the loss-based estimator
is expected to detect the overuse.

BUG=webrtc:7508
==========

to

==========
When receiving an RTCP packet containing feedback about multiple RTP packets,
we currently check for bandwidth overuse once for every RTP packet.

This CL creates an experiment to test processing all packets in the RTCP
feedback before checking for overuse. This can be thought of as checking
for overuse per RTCP packet instead of per RTP packet.

The change is not expected to have a large impact, but enabling the
experiment will make the delay-based BWE slightly less sensitive. This means
that we'll be less likely to back down incorrectly after a brief network
transient, at the cost of sometimes missing real overuse (especially when
the network queues are short). In the latter case, the loss-based estimator
is expected to detect the overuse.

The experiment is off by default.

BUG=webrtc:7508
==========

[terelius, 2017-04-21 10:04:03]

terelius@webrtc.org changed reviewers:
+ michaelt@webrtc.org, stefan@webrtc.org

[commit-bot: I haz the power, 2017-04-21 10:38:45]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2017-04-21 10:38:46]

Dry run: This issue passed the CQ dry run.

[michaelt, 2017-04-24 07:00:52]

lgtm

https://codereview.webrtc.org/2835573003/diff/1/webrtc/modules/congestion_con...
File webrtc/modules/congestion_controller/delay_based_bwe.cc (right):

https://codereview.webrtc.org/2835573003/diff/1/webrtc/modules/congestion_con...
webrtc/modules/congestion_controller/delay_based_bwe.cc:54: std::string
experiment_string =
nit:
how about use "experiment_status" instead of "experiment_string" ?

[terelius, 2017-04-27 11:43:11]

On 2017/04/24 07:00:52, michaelt wrote:
> lgtm
> 
>
https://codereview.webrtc.org/2835573003/diff/1/webrtc/modules/congestion_con...
> File webrtc/modules/congestion_controller/delay_based_bwe.cc (right):
> 
>
https://codereview.webrtc.org/2835573003/diff/1/webrtc/modules/congestion_con...
> webrtc/modules/congestion_controller/delay_based_bwe.cc:54: std::string
> experiment_string =
> nit:
> how about use "experiment_status" instead of "experiment_string" ?

I picked "experiment_string" to be consistent with the code for other
experiments. For example here:
https://cs.chromium.org/chromium/src/third_party/webrtc/modules/bitrate_contr...

[stefan-webrtc, 2017-05-02 07:04:20]

lgtm

[terelius, 2017-05-02 07:47:05]

The CQ bit was checked by terelius@webrtc.org

[commit-bot: I haz the power, 2017-05-02 07:47:11]

CQ is trying da patch.

Follow status at:
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.webrtc.org/...

[commit-bot: I haz the power, 2017-05-02 08:04:27]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1493711225764380, "parent_rev":
"b82ac6aed320ab25c80acc210b20f8d710601d2e", "commit_rev":
"bf2c049a12302141725af0a7407022908fda7efb"}

[commit-bot: I haz the power, 2017-05-02 08:04:30]

Description was changed from

==========
When receiving an RTCP packet containing feedback about multiple RTP packets,
we currently check for bandwidth overuse once for every RTP packet.

This CL creates an experiment to test processing all packets in the RTCP
feedback before checking for overuse. This can be thought of as checking
for overuse per RTCP packet instead of per RTP packet.

The change is not expected to have a large impact, but enabling the
experiment will make the delay-based BWE slightly less sensitive. This means
that we'll be less likely to back down incorrectly after a brief network
transient, at the cost of sometimes missing real overuse (especially when
the network queues are short). In the latter case, the loss-based estimator
is expected to detect the overuse.

The experiment is off by default.

BUG=webrtc:7508
==========

to

==========
When receiving an RTCP packet containing feedback about multiple RTP packets,
we currently check for bandwidth overuse once for every RTP packet.

This CL creates an experiment to test processing all packets in the RTCP
feedback before checking for overuse. This can be thought of as checking
for overuse per RTCP packet instead of per RTP packet.

The change is not expected to have a large impact, but enabling the
experiment will make the delay-based BWE slightly less sensitive. This means
that we'll be less likely to back down incorrectly after a brief network
transient, at the cost of sometimes missing real overuse (especially when
the network queues are short). In the latter case, the loss-based estimator
is expected to detect the overuse.

The experiment is off by default.

BUG=webrtc:7508

Review-Url: https://codereview.webrtc.org/2835573003
Cr-Commit-Position: refs/heads/master@{#17968}
Committed:
https://chromium.googlesource.com/external/webrtc/+/bf2c049a12302141725af0a74...
==========

[commit-bot: I haz the power, 2017-05-02 08:04:31]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/external/webrtc/+/bf2c049a12302141725af0a74...