How to Manage Streaming Latency Expectations
See more videos like this on StreamingMedia.com.
Learn more about streaming latency at Streaming Media West 2021.
Read the complete transcript of this video:
Tim Siglin: Krystal, you've been involved in live for quite some time. And on the broadcast side with partners, Oliver was mentioning networks globally being potentially hostile, but also you've got a lot of legacy devices that you have to deal with, and add things like DRM. What are some of the learnings that you've come away with from doing that kind of work and finding on how decisions are made for what network, or what partners require say in one part of the world versus everybody else in the rest of the world?
Krystal Mejia: The nice thing about being an engineer is you're more focused on how things are done, and so you don't have to focus too much on those business requirements. Those of us on the tech or platform side rely obviously on the brands, the business to tell us what those numbers and metrics are. All we can do is provide the data that says the more components we add to this chain, the more latency is a huge target. They have to understand that this is how much latency we're going to be introducing. And so you look at an event like say the Super Bowl, where we start to make trade-offs and say, "We're going to bypass DAI [Dynamic Ad Insertion] because we don't think we could even scale the DAI service to handle those many concurrents. We have the data to show that, and then we can say, "Okay, we're going to try to figure out how to embed the ads in the stream and generate revenue that way." And so we start to look at other trade-offs, but ultimately, once the business understands their requirements, as an engineer, you can start to look at all those different components and eliminate them from your video pipeline, if needed, or optimize those places and look at redundancy in those places, regardless of if it introduces latency or more latency or not.
Robert Reinhardt: I wanted to ask Krystal a follow-up on that, because you and I wear some similar hats. I do engineering as well. But because I'm a consultant, I wanted to ask, "How often do you find that the people you're working with--whether they're partners who don't understand what their business requirements are, and you have to help inform them, will they come to you with some assertions thinking that they understand it, but they really don't? And what they're asking might be counterintuitive or counterproductive to what they want to achieve.
Krystal Mejia: That happens a lot. We build a platform and we say, "We can do DAI, we can do DRM, we can do everything." And people think we're just plug and play. And they don't understand, on output, what the implications are. And even just on setting up events of that nature. And so, depending on your experience, a lot of people, from a business perspective, just understand data and numbers and don't necessarily understand the technology under the hood. And so we do a lot of work to share knowledge here about the limitations we run up against because the ultimate thing for us is obviously reliability and the quality of service. And so we'll give up on video quality or latency as needed, all depending on the brand and their business requirements.