How QoS Metrics are determined in the Zoom API

Using this template helps us debug your issues more effectively :slight_smile:

Description
It seems like the QoS metrics provided from this dashboard are given at each minute of the call. For metrics like resolution, bitrate, or latency, are the returned values avg/max/median over each minute or are they the stats at the top of every minute (instantaneous) Thanks!

1 Like

Hi @macmillan,

Good question!

I believe these metrics reflect their stats at the top of each minute—however, I’m working on confirming this and will follow up shortly.

Thanks,
Will

Hi Will,

Thanks for the reply. I was also wondering whether there is a log or a way to record the bitrate, resolution, fps, and other QoS metrics, as displayed in the Zoom client. I’m referring to the stats you can see when you open the settings menu during a call.

Thanks!

Hi @macmillan,

At this time, these metrics are only available as returned by the QoS endpoints currently. I’m still working on confirming this other bit for you and will be in touch shortly.

Thanks!
Will

1 Like

Thanks for the response. One other related question: is the latency that’s given the application layer latency or network layer latency?

Hi @macmillan,

I’ve confirmed that our GET Meeting Participant QoS endpoint’s metrics like resolution, bitrate, latency, are returned based on state (e.g., when requested, or last received).

In regards to the latency, I believe this is network layer.

Thanks!
Will

Hi Will,

Thanks for the reply. Can you please tell us how exactly the latency metric is calculated? Assuming, you calculate the network-layer one-way delay for all packets in a video frame, how are these numbers averaged?

On a related note, are there any ways to infer application latency which will also include delay at the sender (packet queuing) and receiver (due to jitter buffer)?

Hey @tmangla,

Let me check with some internal resources on these. I’ll follow up here once I have some confirmation.

Thanks,
Will

Hi @tmangla,

Following up on my previous note, I have some confirmation on this to share.

To that end, the statistical latency is the sum of the upstream and downstream latency, and this time includes the application processing time. The values returned in the API are the 5-second average and the 5-second maximum.

Thanks!
Will

1 Like

Hi @will.zoom

When you say 5-second average and 5-second maximum, is this the average/max latency over 5 seconds at the top of each minute (i.e. 0:00 - 0:05)?

Thanks!

Hi @macmillan,

Let me confirm this :slight_smile:

I’ll follow up shortly,
Will