What's Up with Tech?

Innovating Video Streaming: Edge Nodes, AI-Driven Caching, and Sustainable CDN Solutions

Evan Kirstel

Interested in being a guest? Email us at admin@evankirstel.com

Discover how video streaming is being transformed in the most challenging environments with Steve Miller-Jones, VP of Product Strategy at Netskrt Systems. Learn how Netskirt’s revolutionary approach to Content Delivery Networks (CDNs) is bridging the digital divide in rural areas. By deploying edge nodes within last-mile ISPs and leveraging AI for pre-caching popular content, Netskirt ensures that users experience seamless, high-quality streaming no matter their location. Steve shares the nuts and bolts of their technology, offering insights into how these innovative solutions are reshaping content delivery for communities with poor internet infrastructure.

In this episode, we also delve into the future of connectivity with the integration of 5G and the critical role of sustainability in CDN innovation. Steve explains how Netskrt is optimizing content delivery in diverse settings, from rural ISPs to public transport, while maintaining energy efficiency through the use of low-powered appliances. We explore what differentiates Netskrt from its larger competitors and how their unique philosophy ensures local availability of frequently requested content, minimizing reliance on major CDN servers. To wrap up, we reflect on the profound impact of these advancements and express our heartfelt gratitude to Steve for sharing his visionary insights, as well as to our dedicated audience for their unwavering support.

Support the show

More at https://linktr.ee/EvanKirstel

Speaker 1:

Hey everybody, today we're diving into, ironically, the world of video streaming. Today on a video stream with Netskirt Steve, how are you? I'm?

Speaker 2:

very well. Thank you, evan. Yeah, great day so far. Looking forward to the rest of the week and yeah, as you say, we're being somewhat meta and talking about video on a video.

Speaker 1:

Video and content delivery networks in particular. So maybe some interruptions are in order. If you could introduce yourself your role at Netskirt and what is the big idea? Absolutely.

Speaker 2:

Yeah, I'm Steve Miller-Jones. I'm the VP of product strategy here at Netskirt Systems. Netskirt is a CDN, but a bit more of a specialist CDN than perhaps people might have come across before. We are focused on enabling content delivery at high quality in the hard-to-reach, far-flung reaches of the Internet, and by that I mean locations where people might be a bit further away from towns and cities. They might have a small ISP serving their needs, they might be out in the rural countryside somewhere, so typical locations that maybe so far in the history of the internet have not been well serviced by high speed broadband, and the ISPs that serve those locations sometimes struggle with making their internet services work. So we're here to help them make the best of their networks.

Speaker 1:

Fantastic. So you know, cdns are pretty well understood by many of us in tech. But what's your approach here? I think you have quite a novel approach to enhancing video streaming in these kind of areas with weak or poor internet quality yeah.

Speaker 2:

So our approach is is to think about them, as you know, constellation of highly distributed edge nodes, and we put them into, literally into the last mile subscriber network. So you know, if you sort of peel peel the onion rings of the internet, you know, back in the middle of the internet, somewhere in the cloud, is where most of the connectivity uh, sort of coalesces, if you like, um, and so as you go further out from those that coalesce location, the ways in which you can get to a last mile ISP across the internet are wide and varied and there are lots of carriers, tier one carriers, who are the sort of backbone of the internet, and then lots of intermediate people who are taking traffic from the backbones and spreading it out towards smaller ISPs. Those are typically kind of like internet exchange type locations and then you have your last mile ISP. You know they built this wonderful, you know symmetric gigabits of the home, fiber or great wireless, you know wireless ISP or whatever they might have, and they have great connectivity to the home but upstream from them as they try and get content that's stuck up in the internet in the cloud somewhere or in other CDNs where they're serving from in the middle of the internet. All the congestion happens outside or above the ISP's network and they're not in control of that. And so what we do is we talk to the ISPs about how their network is configured and managed and we deploy a suitably sized cache or a set of caches appliances these are that go into the ISPs network and we configure them to deliver content from a multitude of OTT content providers. And by delivering it from within the network where the subscriber last mile subscriber network is, we're essentially reducing the amount of times we have to go upstream. So we bring the content into the network as few times as possible, we cache it in the subscriber network and we serve all the subscribers from within the last mile network. So we're a very highly distributed specialized CDN.

Speaker 2:

We are for live and video-on-demand content. So you know live is big spikes, large, huge events happening online all the time Well, seemingly all the time. Now I'm sure we've, all you know, read the information about how many rights deals have been done this year for internet. You know OTT delivery, so live is a big thing. But then also VOD right. You know there are huge video on demand libraries out there and everyone's expecting the quality to be the same across the entire library.

Speaker 2:

So how do you best position the content into these last mile networks?

Speaker 2:

So what is likely to be watched is already cached before someone asks.

Speaker 2:

So we have a load of, as you might call, special source, which is orientated around understanding a content provider's library, prioritizing the pieces of VOD content that should go out into any particular location based on usage or based on popularity of that content or based on external influencing factors that we might find out in the internet and look at with machine learning and potentially AI, and we build a profile of what should go into which network and we pre-cache the VOD content so that the user is most likely to get a cache hit from our cache.

Speaker 2:

So we do things a bit differently so that the cache becomes useful. An empty cache CDN cache in an ISP network is fairly useless. A CDN cache that we've been able to pre-populate with the VOD content that's actually popular and is going to be or most likely to be watched in that network location becomes a valuable asset to the ISP and delivers high-quality QOE metrics back to the content owner. So there's a few complicated pieces in there, but it's a bit of a different take on how to solve the problem of getting to everyone on the Internet.

Speaker 1:

Wow, it's an interesting challenge, to say the least, and I see here you're using AI and machine learning to improve content delivery.

Speaker 2:

How does?

Speaker 1:

that work exactly.

Speaker 2:

Yeah, I touched on it briefly that we essentially use it for understanding popularity of a video-on-demand library, so that we can fill our caches and pre-position content effectively, and so we look at trend in usage.

Speaker 2:

We also look at external data sources, including the content provider's own apps, how they lay out their content, what the content provider's own their apps, how they lay out their content, what the content looks like online, what we're getting out of your ratings rankings, what you might find being talked about online, and consuming a broad set of data in order to understand. Well, we know that this piece of content is being positioned and promoted on the first page, like the top rail of your OTT library. Is it really popular, right? Or should we just take the content provider's word for it that that's what's going to get watched? And so our AI ML work is there to improve that scoring of content. So we build a stack, ranked set of prioritized content for pre-caching. So you know it's interesting. And, yeah, we can take many external sources, but building on utilization, usage and expected consumption habits is what we do mostly.

Speaker 1:

Interesting. So we've seen with 5G kind of the whole video market sort of turbocharged, which has been great for consumers, businesses. You know what role does 5G play in enhancing Netskirt services and I guess vice versa, how do you enhance, you know, the 5G experience?

Speaker 2:

Yeah, so there's a couple of ways that Netskirt has played in the 5G space. In addition to ISP caches and putting our CDN into ISPs, we have also worked in putting our CDN into public passenger transport environments. So we've had caches on passenger rail trains in the UK and those are essentially trains that are connected with four LTE or 5G, and then you're sort of trying to make the best use of that somewhat inconsistently available connectivity. If you've ever ridden a train, you know they go through tunnels, right, so you lose your connectivity. And so how do you make the best use of that 5G connection? How do you fill up a cache with stuff that is likely to be watched? And so we have a bunch of tools that allow us to make use of connectivity when it's available. So if we come into a connectivity-rich area we can do a lot of activity quickly to download and preposition or cache a lot of stuff that has been requested that we didn't have in cache yet or is just being published, and we have a new set of stuff to push in.

Speaker 2:

And additionally, you know we work with a lot of wireless ISPs. So you know WISPs across the UK and some in the States. You know whether or not. The 5G piece is what exactly they're deploying. I mean, basically it's giving a higher rate of bandwidth depending on the type of deployment, and so deploying caches into some parts of the wireless ISPs again gives them the benefit of being able to distribute content from within their network without having to send everything upstream. So you know they're not just using it as a pass-through, they're beginning to make a good use of their network. So you know 5G has its place in the transportation use cases and in addition to trains we're in airplanes sometimes as well. But it has a play in ISPs out in the rural wilds as well, who are relying on wireless more and more.

Speaker 1:

Fantastic. A lot of focus now on sustainability, energy efficiency, both in the data center and the edge. You know video is, of course, a big user of those resources. How do you think about content delivery and energy efficiency and those sort of topics?

Speaker 2:

Yeah, it's definitely an interesting topic the eternal question of if someone adds another cache, does one get turned off somewhere else. So I think it's important, when we think about electricity consumption and sustainability, to be thinking about something that's consequential, rather than just just measuring carbon usage or trying to attribute it to a certain part of the workflow. For us, with the isp work that we're doing, we're providing a particular type of appliance, and our goal with these appliances is to be relatively low powered, so they don't have huge numbers of fans, and we look carefully at what level of hard drive we put into them, what CPUs we're using so to regulate for heat a bit more, and so what we're trying to do is identify appliances that can be relatively low power compared to just a customer off the shelf. You know one new server from insert manufacturer here. So we have a focus on providing an appliance to the ISP which is designed to be a little bit less power hungry, and so that's an approach we've begun with.

Speaker 2:

We're a startup and we'll be looking at ways in which we can, you know, hopefully reduce the power consumption needed for small caches over time. Yeah, there's a new generation of CPUs going to come out that again maybe require less fans and if you can reduce your spinning fans, you're going to reduce your power consumption. So that's the sort of starting point for us. Assumption. So that's the sort of starting point for us, and I think you know there's there's a lot to come in the sustainability side of you know all the cloud providers, all the CDNs, everywhere. You know who's carrying the traffic and did anyone switch anything off because someone else is carrying the traffic. So I think there's a lot of interesting discussion in that Our goal is to make our footprint as light as we can at this point in time and accessible for the ISPs and for the content providers.

Speaker 1:

Okay, well done. As a startup in this space, you know better than I. There are some very big companies in this space 800-pound gorillas, I think of Akamai down the road here in Cambridge. What you know sets you apart from other players in the CDN space, and how do you see yourself?

Speaker 2:

You know, I think the thing that sets us apart is the philosophy of where we've come from and why we're trying to solve the problem we are, and the kind of problem we're trying to solve, the the founders talk about. Um, you know, they met the hotel somewhere in middle of nowhere one day and they were highly frustrated that they were not able to, uh, watch what they wanted to watch in their hotel room using their own OTT accounts. And they started talking about well, surely there must be a better way than everyone in this hotel making a request for a similar range of content. Most likely, you know, maybe 60% of the stuff that can be watched is the same across three or four platforms. I don't know if that number's real, but that's the kind of number. Does it all have to go upstream? Does it all have to go to the nearest? You know, you said Akamai. Where's the nearest Akamai server?

Speaker 1:

And they're you know, you said Akamai, where's the nearest Akamai server?

Speaker 2:

And they're, you know, they're not necessarily individuals, they're in data center locations Where's the nearest? Or to any of the other CDNs. And so they started thinking, actually there's a benefit to locations that are somewhat isolated or in the outskirts of the internet to override the fact fact they have to go upstream and begin to provide a mechanism where they can get the content locally. And so it the the approach for us is different because we, rather than have a dense set of hardware in specific locations, we have a highly distributed uh constellation of edges all over the place, and our challenge to ourselves and to present this up to a content provider to use is how do we make this some kind of unified control plane that is verified access to a validated set of users in a specific location? So that's what we're doing, which is somewhat different from your other CDNs. Your other CDNs are connected to the internet through a variety of direct connections, peering to one, two, three, transit, and they're going to say well, of course, I can reach that ASN, I can reach the entire internet from this location, but to get to a smaller set of consumers in a specific ISP that maybe is not a large national provider, maybe a small, regional, local ISP? Do they know actually what the path is? Can they verify that they have actual access to that network and what the results in that network of their quality metrics are?

Speaker 2:

And what've found is that these smaller isps they're all the sort of 20 and under population numbers, right, like 80 percent of the cdn population is in big urban areas, right, so they get millions of people in new york city, okay, watching, watching oct all the time, and that is the majority of the QOE metrics for big CDM providers. Smaller ISP in upstate New York and Buffalo is going to be serving a much smaller set of people, right, and their metrics are just as important, but they're just much smaller. So they're statistically insignificant to your normal CDM. And we make those people statistically significant. So we're bringing what are what look like outliers in your data set for most content owners. We're bringing those into focus and saying these are your subscribers just as much as anyone else. Let's make sure they're getting what they expect and what you believe you can give them by again you know, verifying and validating that we have access to that network not just saying of course we can reach it.

Speaker 2:

We're on the internet and not really knowing what happens when I send the bits. We know what happens when we send the bits.

Speaker 1:

Fantastic approach, the startup. What kind of partnerships are you building with ISPs or transport companies or others to serve these customers or transport companies or others to serve these customers.

Speaker 2:

Yeah, I mean, our partnerships over the years have been really important to us. We've partnered with a company called Talus for in-flight entertainment, in-flight communication. We've been partnered with a company called Bolden for some of the transport infrastructure pieces as well. We go directly and talk to ISPs across the usa, canada, uk, italy and brazil, and so we work in some of the isp associations uh, the whisper group in the states. Yeah, expanding that to try and make sure that you, we can tell the story that we believe has value for the ISP and the content owner to them, but also that we're listening to what they need right, because you know, it's always a two-way conversation.

Speaker 2:

The ISPs have got a range of different architectures that they deploy with and they have slightly different needs to each other, and we've adapted our product to fit different types of network deployments. And then the content owners themselves, the OTT platforms that we all subscribe to, know and love. Well, they are huge partners for us. They are concerned with ensuring everybody gets their content. They all have particular ways of deploying their services. The kpis might be very similar, but they'll. They'll have nuances in how those services work, and so we take part in industry groups like the, the svta, the streaming video technology alliance, as well, and so you know, we sort of keep ourselves across the active bodies in the internet environment that are trying to solve well, how should we best do this? And we're trying to keep ourselves across the groups who are ISP customers, who are actually saying this is what we need for our networks to be successful. So it's a good combination of content one hand and access to subscribers in the other hand. And how do you best put them together?

Speaker 1:

Fantastic approach. So what's up for the rest of the year? Any plans for new industries, new regions, markets? What's on your mind for the remainder of the year?

Speaker 2:

Yes, we've got a pretty busy season happening now, so stateside it's the new NFL season, so that's busy for us. There's some excitement around hockey as well for us this year, which is going to be cool. So we're busy making sure the capacity is in the right places. We've begun to deploy into the UK and into Italy and we're looking at how best to deploy to ISPs in Brazil. So that keeps us busy with different environments for different languages and people behave and operate in slightly different ways and the value change are slightly different in different places.

Speaker 2:

I'm off to IBC this week, so I'm heading out to the great city of Amsterdam for IBC, which is the culmination of lots of content owners, content providers, platforms, fast channels all of the technology you could ever imagine is needed for end-to-end video delivery, so that's going to be super cool. And there's a SVTA membership meeting as well in Antwerp next week. We'll be going to that and then we're doing a few other things over the end of the year with the ISP conferences and then we'll be looking towards NAB in the States early next year from a conference perspective as well. But you know, got our heads down. We're trying to ensure that we're delivering a good service for our content partners, and so the most important thing is that, of course every day that the quality metrics flow in the right direction.

Speaker 1:

On that note, a mic drop moment, we'll say farewell. Thanks so much for the insight and the sharing, your vision and mission and it's an important one. Really great work and thanks for joining.

Speaker 2:

No problem, evan, it's been an absolute pleasure. Thank you very much.

Speaker 1:

Thank you and thanks everyone for listening and watching, until next time.