Event banner
AMA: Microsoft Teams Performance and Overview
Event details
I can help here, which MediaLine_ fields in particular are you interested in knowing more about?
The 'Other' section in the Advanced tab (as well as the Debug tab) are both primarily geared towards internal engineering teams. We use them for investigating application behavior and assisting with customer incidents. Many of those fields won't be particularly useful to customers on their own, while some of those fields are found in Call Quality Dashboard (albeit under more friendly names with the 'MediaLine_' bits truncated). For those that are available in CQD, we have definitions available here: Dimensions and measurements - Call Quality Dashboard (CQD) - Microsoft Teams | Microsoft Docs
- Petri-XApr 28, 2022Bronze Contributor
That would be awesome if you could open even a bit these mysterious 🙂
And yes, I believe these are mostly used for internal purposes, but when someone is asking what went wrong, we need to turn all of our rocks.
The following lines has been lately my interest:
MediaLine_SystemMetric_SystemCpu_CpuFrequencyHistogram
MediaLine_OutboundStream_Payload_Video_VbssTypeHistograms
MediaLine_InboundStream_Payload_Audio_RecvSignalLevelHistogram1
MediaLine_OutboundStream_Payload_Audio_SendSignalLevelHistogram1
MediaLine_InboundStream_Payload_Audio_NetworkJitterHistogram1
MediaLine_InboundStream_Payload_Video_RecvDelayEngineHist
MediaLine_InboundStream_Payload_Video_RecvDelayTransportHistMediaLine_InboundStream_Payload_Audio_JitterBufferResetCount
From those lines I would like to try to see if I can get any ideas how the call has been went for user point of view.
Also would be nice to know how to turn the timestamps to normal time e.g. 38272845477884304. As those seems to not be Epoch?
Are you also a person who understand the content of Advanced very well? I had a case also, but that did not work well. I believe some of the values are incorrectly written on there. Eg. If you take section "inbound network" and from there e.g. Maximum round-trip time. And then compare that to "outbound network". The values are in same order e.g.
Inbound Network Max Round Trip: 49ms | 59ms
Outbound Network Max Round Trip: 49ms | 59ms
But then e.g. Maximum Jitter, the values are mirror of others:
Inbound Network Max Jitter: 3ms | 4ms
Outbound Network Max Jitter: 4ms | 3ms
. . .I have so many questions... 😄
Like on the same case I had a feeling that outbound and inbound were mixed elsewhere as well.
- JamesParkesMSFTMay 02, 2022
Microsoft
I'll start with the 'Histogram' fields. Histogram fields start with the total number of Observations, followed up by semicolon-delimited sets of three values: LowerBound, UpperBound, and Occurences. Here's a sample of a MediaLine_OutboundStream_Payload_Audio_SendSignalLevelHistogram1 value: 179988;-90,-87,2;-87,-84,142;-84,-81,1496;-81,-78,831;-78,-75,3477;-75,-72,7125;-72,-69,7552;-69,-66,4530;-66,-63,3205;-63,-60,3028;-60,-57,3280;-57,-54,3365;-54,-51,3841;-51,-48,4549;-48,-45,5223;-45,-42,6406;-42,-39,7877;-39,-36,9972;-36,-33,12554; 47285;Min,-96,3;-96,-93,1;-93,-90,2;-90,-87,8;-84,-81,57;-81,-78,39008;-78,-75,6587;-75,-72,316;-72,-69,195;-69,-66,152;-66,-63,111;-63,-60,98;-60,-57,62;-57,-54,49;-54,-51,55;-51,-48,37;-48,-45,40;-45,-42,36;-42,-39,27;-39,-36,28;-36,-33,30;-33,-30,43; The first value, 179988, is the total number of Observations for the line. This is the only time in that line you'll see the total observations. Where a histogram metric has two lines (like this one), you'll see it starts with another total number of observations for that line. In this example, the second line has 47285 observations of values in that line. Next are the three values between the semicolons. -90 is the LowerBound -87 is the UpperBound 2 Occurrences. Then the next set: -87 is the Lowerbound, -84 is the UpperBound, and 142 Occurrences. And so on. If you see a 'Min' or a 'Max', that is the minimum or maximum possible value for the particular metric being measured. Our data scientists in media and call quality use these histogram values to help build models to better understand various network, audio, and device conditions that can contribute to good and not-so-good calling experiences. I can check with engineering about the timestamps and those values being flipped or not flipped. Hope this has provided some clarity!