Video: The 8 Key Elements of Live Stream Monitoring
Learn more about live stream monitoring at Streaming Media's next event.
Watch the complete video of this presentation from Streaming Media West, DT201A: Non-Reference Picture Quality Assessment in the Streaming Media Conference Video Portal.
Read the complete transcript of this clip:
Andrew Scott: I’m describing some of the key elements of a monitoring tool. I won't read from the slides. These are available for you to download as well. But I'll highlight some of the key points here. First, I'll talk about the difference between QoS and QoE. So I think it's important to look at both.
QoS (quality of service) is, essentially, how well my content is being delivered. It's about the delivery, as opposed to QoE, quality of experience, which is more about the content. We can have viewer issues due to QoS, such as download errors from the fragments.
A QoE issue might be resulting from encoding at a too-low bit rate. So you've got a high amount of compression artifacts. You don't have a good quality of experience for the viewer even though the delivery is fine. The encoding is fine; it's just set up to not provide as clean a picture as perhaps we could have delivered to them.
Here’s another example talking about API control. This is actually something that Billy's going to touch on later, what they've implemented at Fubo.
A key element of monitoring is the ability to scale alongside with what the service is that you're delivering. If you've got 100 live streaming channels, each of those is many representations in the ABR bitrate ladder. We need our monitoring tool to be able to look at all of that simultaneously, aggregate that up, and bring it back to a usable dashboard. And a key part of that is an API where you can pull the necessary metrics that you want, and basically play with that data, and make it available in a way that's more meaningful for your operators.
A great example of where they need to bring some of the stats from the monitoring tool and correlate those with other pieces of equipment. The way of doing that is a unified dashboard where you've got talking to the API of multiple devices.
What's another good example? Talking about thresholds and alerts. It's possible for a monitoring system to produce too much data. From talking about that scale of hundreds of channels and many representations on each channel, that's the possibility of generating a tremendous amount of data.
It's useful, sometimes, to be able to go back and look at a moment in time. We had viewer complaints from the program that was airing from 8:00 o'clock to 9:00 o'clock. We had some issues.
You can go back and pull that data out of the database, and take a look at what was happening at that time. So it's useful to save all of that data, but you don't necessarily want to bombard an operator with a lot of issues.
That's where a good threshold and alerting system comes into play. The ability to set at what point I want to send a notification to someone, whether it's an email, or a Slack message, or something like that.
Interra Systems VP Product Management Anupama Anantharaman offers essential recommendations for effective streaming quality and performance monitoring in this clip from her presentation at Streaming Media West 2019.
Interra Systems VP Product Management Anupama Anantharaman definitively answers the question "Why monitor?" for OTT providers in this clip from her presentation at Streaming Media West 2019.
Tektronix Video Product Manager Dan Murray identifies the eight essential elements of live stream monitoring in this clip from his presentation at Streaming Media East.
Tektronix Video Product Manager outlines the efficiency gains of cloud-based streaming and raises the question of how those advantages enhance video quality in this clip from his presentation at Streaming Media East.
According to Bitmovin's Sean Longworth, it depends.
Yorktel's Joe Arena makes the case for on-prem encoding, storage, and management when choosing an enterprise video platform, and having a strong support model in place.