|
|
Created:
4 years, 3 months ago by Sergey Ulanov Modified:
4 years, 3 months ago Reviewers:
Irfan CC:
chromium-reviews, chromoting-reviews_chromium.org Target Ref:
refs/pending/heads/master Project:
chromium Visibility:
Public. |
DescriptionMeasure bandwidth utilization in the ScrollPerformance test
Now ScrollPerformance also reports network bandwidth utilization.
Also removed two outdated tests: StreamFrameRate and IntermittentChanges.
These tests were intended mainly to evaluate performance of the PseudoTCP
layer and they were using verbatim video codec. They are not applicable
for WebRTC-based protocol. Also the protocol perf tests have been updated
to use SoftwareVideoRenderer which allows to get FrameStats when using
ICE protocol and this approach is not compatible with the old tests.
Bandwidth utilization numbers I get currently with 8Mbps bandwidth:
ICE: 93%
WebRTC: varies from 21 to 25%
The second number also goes to 32 when I change test length and warm-up
time to 5 seconds. This suggests that the problem is partly that the
bandwidth estimate in WebRTC is below the actual bandwidth in the test and
it takes a lot of time to ramp up.
BUG=645656
Committed: https://crrev.com/d5ce2225173ec41804750d3de0211408ba096bb9
Cr-Commit-Position: refs/heads/master@{#418444}
Patch Set 1 : . #Patch Set 2 : . #
Total comments: 4
Patch Set 3 : units #Patch Set 4 : header #Messages
Total messages: 18 (9 generated)
Patchset #1 (id:1) has been deleted
Description was changed from ========== Measure bandwidth utilization in the scroll perf test BUG=645656 ========== to ========== Measure bandwidth utilization in the ScrollPerformance test Now ScrollPerformance also reports network bandwidth utilization. Also removed two outdated tests: StreamFrameRate and IntermittentChanges. These tests were intended mainly to evaluate performance of the PseudoTCP layer and they were using verbatim video codec. They are not applicable for WebRTC-based protocol. Also the protocol perf tests have been updated to use SoftwareVideoRenderer which allows to get FrameStats when using ICE protocol and this approach is not compatible with the old tests. Bandwidth utilization numbers I get currently with 8Mbps bandwidth: ICE: 93% WebRTC: varies from 21 to 25% The second number also goes to 32 when I change test length and warm-up time to 5 seconds. This suggests that the problem is partly that the bandwidth estimate in WebRTC is below the actual bandwidth in the test and it takes a lot of time to ramp up. BUG=645656 ==========
sergeyu@chromium.org changed reviewers: + isheriff@chromium.org
https://codereview.chromium.org/2338693002/diff/40001/remoting/test/protocol_... File remoting/test/protocol_perftest.cc (right): https://codereview.chromium.org/2338693002/diff/40001/remoting/test/protocol_... remoting/test/protocol_perftest.cc:214: EXPECT_GE(frame_stats_.size(), num_expected_frame_stats_); should this be EXPECT_EQ ? https://codereview.chromium.org/2338693002/diff/40001/remoting/test/protocol_... remoting/test/protocol_perftest.cc:581: VLOG(0) << "Average latency: " << latency_sum.InMillisecondsF() / num_frames; Output units on latency and total size ?
https://codereview.chromium.org/2338693002/diff/40001/remoting/test/protocol_... File remoting/test/protocol_perftest.cc (right): https://codereview.chromium.org/2338693002/diff/40001/remoting/test/protocol_... remoting/test/protocol_perftest.cc:214: EXPECT_GE(frame_stats_.size(), num_expected_frame_stats_); On 2016/09/13 20:19:15, Irfan wrote: > should this be EXPECT_EQ ? No. Potentially multiple VideoFrameStats messages may be received in bulk. We cannot guarantee that no new FrameStats received after OnVideoFrameStats() calls RunLoop::Quit() and before RunLoop::Run() returns. https://codereview.chromium.org/2338693002/diff/40001/remoting/test/protocol_... remoting/test/protocol_perftest.cc:581: VLOG(0) << "Average latency: " << latency_sum.InMillisecondsF() / num_frames; On 2016/09/13 20:19:15, Irfan wrote: > Output units on latency and total size ? Done.
lgtm
The CQ bit was checked by sergeyu@chromium.org
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...
The CQ bit was unchecked by commit-bot@chromium.org
Try jobs failed on following builders: win_clang on master.tryserver.chromium.win (JOB_FAILED, http://build.chromium.org/p/tryserver.chromium.win/builders/win_clang/builds/...)
The CQ bit was checked by sergeyu@chromium.org
The patchset sent to the CQ was uploaded after l-g-t-m from isheriff@chromium.org Link to the patchset: https://codereview.chromium.org/2338693002/#ps2 (title: "header")
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...
Message was sent while issue was closed.
Description was changed from ========== Measure bandwidth utilization in the ScrollPerformance test Now ScrollPerformance also reports network bandwidth utilization. Also removed two outdated tests: StreamFrameRate and IntermittentChanges. These tests were intended mainly to evaluate performance of the PseudoTCP layer and they were using verbatim video codec. They are not applicable for WebRTC-based protocol. Also the protocol perf tests have been updated to use SoftwareVideoRenderer which allows to get FrameStats when using ICE protocol and this approach is not compatible with the old tests. Bandwidth utilization numbers I get currently with 8Mbps bandwidth: ICE: 93% WebRTC: varies from 21 to 25% The second number also goes to 32 when I change test length and warm-up time to 5 seconds. This suggests that the problem is partly that the bandwidth estimate in WebRTC is below the actual bandwidth in the test and it takes a lot of time to ramp up. BUG=645656 ========== to ========== Measure bandwidth utilization in the ScrollPerformance test Now ScrollPerformance also reports network bandwidth utilization. Also removed two outdated tests: StreamFrameRate and IntermittentChanges. These tests were intended mainly to evaluate performance of the PseudoTCP layer and they were using verbatim video codec. They are not applicable for WebRTC-based protocol. Also the protocol perf tests have been updated to use SoftwareVideoRenderer which allows to get FrameStats when using ICE protocol and this approach is not compatible with the old tests. Bandwidth utilization numbers I get currently with 8Mbps bandwidth: ICE: 93% WebRTC: varies from 21 to 25% The second number also goes to 32 when I change test length and warm-up time to 5 seconds. This suggests that the problem is partly that the bandwidth estimate in WebRTC is below the actual bandwidth in the test and it takes a lot of time to ramp up. BUG=645656 ==========
Message was sent while issue was closed.
Committed patchset #4 (id:2)
Message was sent while issue was closed.
Description was changed from ========== Measure bandwidth utilization in the ScrollPerformance test Now ScrollPerformance also reports network bandwidth utilization. Also removed two outdated tests: StreamFrameRate and IntermittentChanges. These tests were intended mainly to evaluate performance of the PseudoTCP layer and they were using verbatim video codec. They are not applicable for WebRTC-based protocol. Also the protocol perf tests have been updated to use SoftwareVideoRenderer which allows to get FrameStats when using ICE protocol and this approach is not compatible with the old tests. Bandwidth utilization numbers I get currently with 8Mbps bandwidth: ICE: 93% WebRTC: varies from 21 to 25% The second number also goes to 32 when I change test length and warm-up time to 5 seconds. This suggests that the problem is partly that the bandwidth estimate in WebRTC is below the actual bandwidth in the test and it takes a lot of time to ramp up. BUG=645656 ========== to ========== Measure bandwidth utilization in the ScrollPerformance test Now ScrollPerformance also reports network bandwidth utilization. Also removed two outdated tests: StreamFrameRate and IntermittentChanges. These tests were intended mainly to evaluate performance of the PseudoTCP layer and they were using verbatim video codec. They are not applicable for WebRTC-based protocol. Also the protocol perf tests have been updated to use SoftwareVideoRenderer which allows to get FrameStats when using ICE protocol and this approach is not compatible with the old tests. Bandwidth utilization numbers I get currently with 8Mbps bandwidth: ICE: 93% WebRTC: varies from 21 to 25% The second number also goes to 32 when I change test length and warm-up time to 5 seconds. This suggests that the problem is partly that the bandwidth estimate in WebRTC is below the actual bandwidth in the test and it takes a lot of time to ramp up. BUG=645656 Committed: https://crrev.com/d5ce2225173ec41804750d3de0211408ba096bb9 Cr-Commit-Position: refs/heads/master@{#418444} ==========
Message was sent while issue was closed.
Patchset 4 (id:??) landed as https://crrev.com/d5ce2225173ec41804750d3de0211408ba096bb9 Cr-Commit-Position: refs/heads/master@{#418444} |