Voices of Video
Explore the inner workings of video technology with Voices of Video: Inside the Tech. This podcast gathers industry experts and innovators to examine every facet of video technology, from decoding and encoding processes to the latest advancements in hardware versus software processing and codecs. Alongside these technical insights, we dive into practical techniques, emerging trends, and industry-shaping facts that define the future of video.
Ideal for engineers, developers, and tech enthusiasts, each episode offers hands-on advice and the in-depth knowledge you need to excel in today’s fast-evolving video landscape. Join us to master the tools, technologies, and trends driving the future of digital video.
Voices of Video
Beyond CDNs: Conversations on Infrastructure Scaling with i3D.net
Stefan Ideler, co-founder and Chief Innovation Officer of i3D.net, takes us behind the scenes of what makes a truly resilient global network. Founded in 2002, i3D.net has positioned itself as "the scaling experts" rather than just another CDN, working with gaming giants like Electronic Arts, Ubisoft, and Discord to ensure their services scale seamlessly during crucial launches and beyond.
What truly distinguishes i3D.net becomes clear as Stefan explains their network philosophy. While many providers prominently advertise their number of Points of Presence (POPs), Stefan reveals that the magic lies in how these points connect and who they connect with directly. With over 9,000 direct peering relationships - making them the second most interconnected network globally - i3D.net carries customer data much further into the delivery chain than competitors who quickly hand off to third parties.
This approach was born from serving the most demanding users on the internet: gamers. As Stefan colorfully puts it, "the controllers are already through the window if there's only a small interruption." By maintaining control and visibility over traffic from origin to the ISP's doorstep, i3D.net can immediately diagnose and address issues rather than waiting days for information from multiple middlemen.
Stefan also offers invaluable wisdom about avoiding cost traps when scaling infrastructure. He cautions against being seduced by initial incentives like free credits or the convenience of all-in-one cloud solutions. Instead, he recommends envisioning your needs when successful at scale and building a hybrid architecture that balances dedicated hardware (via their "Flex Metal" solution) for predictable workloads with true cloud flexibility only where needed.
Want to see how i3D.net's approach could transform your streaming infrastructure? Visit them at the NETINT booth during NAB or reach out directly to discover how the scaling experts can help your service deliver consistently excellent experiences, even when the internet itself is breaking.
Stay tuned for more in-depth insights on video technology, trends, and practical applications. Subscribe to Voices of Video: Inside the Tech for exclusive, hands-on knowledge from the experts. For more resources, visit Voices of Video.
voices of video. Voices of video.
Speaker 1:The voices of video voices of video well, welcome to another special nab episode of voices of Now. If you're listening to this after the 2025 NAB show, make sure you don't tune off, because this is an episode that you definitely want to hear. What our guest is doing, what they built and it might be applicable to your business, but we are talking to various companies across the VPU ecosystem, learning what they're doing, how they're using VPUs and really having a great time. So today I am joined by Stefan Edler, who is the co-founder of i3dnet. Stefan, welcome to Voices of Video. Thank you, mark. I'm very happy to be here. Yeah, well, it's great to have you and you know I'm not going to steal your thunder because you know we're really excited about what you're doing and I think a lot of our listeners are going to want to hear and learn more. But you know, for those who don't know who the company is, why don't you introduce yourself and i3dnet?
Speaker 2:Thank you. So yeah, I'm Stefan. I'm the co-founder and chief innovation officer at i3dnet, joined the company more than 20 years ago and we've been working with games, streams and all kinds of internet problems ever since. So to give you a quick rundown on what i3D Internet actually does, the company was founded in 2002. And we see ourselves as the scaling experts. So over all those years we've worked with companies like EA, ubisoft or Psyonix or Discord to provide them the best possible scaling experience when it was needed for the applications.
Speaker 2:Of course, when you look at gaming, you have the development cycle of the game, where the infrastructure needs are relatively stable, where the infrastructure needs are relatively stable.
Speaker 2:But then when the game gets released and if the game is a success, the sky is the limit and you need to work with a partner who can take away all of your worries about the whole infrastructure and network aspects of scaling that capacity for your game or any application.
Speaker 2:Really, we do so by truly controlling the full stack of the whole infrastructure. It starts with the data centers to the actual machines and all of the network connectivity between those data centers and towards the internet providers of all the gamers or consumers, how you want to call them. We do so by having hands-on, 24-7 expertise. You talk to real people with real knowledge to really help you and give you insight when you call with us or speak with us or text us on Slack, for example, and we've created many bespoke solutions to make sure that the experience of your game session or stream session will remain uninterrupted by attacks from bad actors, will be uninterrupted by fiber cuts or subsea cables getting disrupted, because for us, the experience of the user is what makes or breaks the scaling journey of your product, and we believe we are not just a mere infrastructure provider. We are the partner for the scaling companies.
Speaker 1:Yeah, yeah, I love that. I like how you even see yourself not in your introduction there about the company. I didn't hear the words content distribution network or you know, or now you know clearly that's what you're doing, right, but it's, it's scaling and I can't think of a more demanding consumer than a gamer. Correct, a demanding customer would be a cable subscriber who, on the Sunday or the Saturday, football game cable goes out. But the point is all jokes aside, it really is the focus on the consumer that I like. So let me ask you you've been at this since 2002. Are you exclusively carrying or distributing video, you know? Are you also involved in like game downloads and kind of large object delivery? Or maybe you can explain, you know what types of content that your network supports?
Speaker 2:So the origins are definitely game-related traffic. So when people were playing games online together with your friends, the traffic from those gamers towards the game servers is something that we would provide and we would also have that game infrastructure running in all of our data centers. Of course, over the years it expanded to also offering the patches for the games, some of the content distribution regarding that as well. So when we look at ourselves today, there's definitely a big content distribution part. There's a big part which ensures for video and voice streaming and calls, and then there's a big part which is having the actual infrastructure where your applications run on, including the applications who stream, of course, towards your clients. So it's a combination of these different aspects and because we have all the different layers of the infrastructure under our control, that allows us to combine these three into an effective solution. Yeah, interesting.
Speaker 1:So you referenced that Electronic Arts is, or has been, I guess, one of your customers. Are you still distributing traffic for Electronic Arts?
Speaker 2:We are still working with Electronic Arts.
Speaker 1:Yeah, and Ubisoft, but Discord is the one that I think a lot of our listeners would be familiar with, possibly even as users. I mean, discord is a dominant platform, largely for gamers, but it's really expanding, right? So, yeah, so you know, what I'd like to understand and this connects to a question that you know that I think is relevant, we were discussing this before we hit record is that you're coming from. You know and these are my words, but you know, you're coming at the sort of traditional video streaming market from a little bit of a different direction, and I view that as a very positive thing because it causes you to look at the challenges of streaming, you know, in a different way, right?
Speaker 1:You know when I think about it, it you know it's easy to assume oh, cdn, there's a standard set of technologies, there's common approaches. You know company A, company B, company C, they're all fundamentally the same, except for you know, company A, company B, company C, they're all fundamentally the same, except for you know the number of pops they have. You know one might have slightly better coverage in LATAM and another one has better coverage in North America, and you know. But they're fundamentally the same. But you and I know that's not true. So yeah.
Speaker 1:So so maybe you can give us, you know, a bit of an introduction, or you know, a shallow but deeper dive into the technology that you've built and what you're bringing to this traditional streaming video market that you know may not be in every platform. Sure.
Speaker 2:So you're totally right there. There's definitely a lot of suppliers in this space of offering content distribution, video and voice streaming, and while from a distance they might all look the same when you zoom in.
Speaker 1:Their websites all use the same words Exactly.
Speaker 2:They all use the same words and then, when you zoom in, you will start noticing the difference. You might see that, oh, some are actually reselling a white label product from another provider. Some might have their own infrastructure, some might have parts of that infrastructure. Some might have transparent pricing and some might have very untransparent pricing based on where your traffic goes in the region, and some might have very untransparent pricing based on where your traffic goes in the region.
Speaker 2:And the one that is to me often a little bit forgotten is how do they actually run and manage their networking infrastructure? So some of the companies might have a lot of points of presence around the world, but it could very well be that they might all be, as each point of present might be their own little little network where and it's not all interconnected together, which is fine for quite a few applications, don't get me wrong. But when you have applications which also depend on connectivity between regions, or where your audience is a little bit more broad, or when you're making calls with your friends across the world, having that single big global backbone starts to be of more importance. Secondly, the Internet itself is not optimized. It breaks all the time.
Speaker 2:It's not just inside random places. Cables get cut. We see it with the subsea cables nowadays in Europe A lot of news, of course, recently. It is a very unpredictable beast. You need to put a lot of control on that and you can do so if you do have your own connections and pops all across the world to be able to work around the problems which are constantly occurring. The internet connectivity might break for 30 seconds, 60 seconds or a few minutes, which traditionally for most office applications is not that big of a problem. You refresh a few times and you go on with your life. But in gaming the controllers are already through the window if there's only a small interruption because the game session will crash and all the players will be disconnected and there will be rage on the forums and on X and FOL.
Speaker 1:Like I said, there's nothing like a gamer, customer base, audience to elicit feedback.
Speaker 2:Yes. So what we see is that there's two main ways to see how people solve this problem. One solution is that every bit and byte that is generated in the data center by the POP, they give those bits and bytes to a third-party provider as soon as possible. That is to limit the costs, because if you don't have to transport all the bits and bytes yourself, then it's obviously a bit cheaper to just give it away and then you just focus on getting as much volume as you can with those third-party providers and then you have a bare bones ready to offer a CDN or voice and video product. But it works fine until things go wrong, because then you have your client calling you. Like, my subscribers in the Western part of the US have trouble accessing the content quickly. Can you see what's going on? Because you are not serving that traffic yourself, but your third party is. We might have outsourced it to another sub-party in that specific region. It might take you days or weeks to get information back to your client, to actually tell them what is going on. So what we see is that other parties and we do the same is we try to carry the bits and bytes as much as we can, as far as we can, towards the internet service provider of the consumer. And then when we finally reach that provider, then we exchange the bits and bytes because then we can tell to our clients and also to the angry gamers on X. We carry your bits and bytes all the way from, for example, Rotterdam, europe, towards Dallas in the US and we know exactly what happened to it. So if something goes wrong we know also exactly. You can tell them right away what went wrong. And we strongly believe, coming from the gaming space, that that's the most efficient way to create quality for your consumers.
Speaker 2:And we also see this in streaming and especially live streaming as well, because for live streaming you want that predictability. You also definitely want low latency, because if you click play on your video and it's in there within half a second, whereas with the competitor it would take five seconds or worse to finally buffer in all the throughput. That's a bad experience. No, it needs to be snappy. You want it to be stable, because the buffering is constantly going on.
Speaker 2:You want an 8K video stream to actually stay 8K at your consumer and not suddenly downgrade to 480p and square on their 80-inch television. That's a bad experience. So you want predictability. So you want, as a provider, you want the visibility and the constant to be able to give a guarantee to the clients and just also to the consumers that that is in good order. But what I just told you you will not see written on the can on many of the websites, or they might write it but then don't have it. So I can definitely understand why it can be very, very difficult to find and pick the right provider.
Speaker 1:Yeah, you know I have worked in the streaming industry around video distribution, video streaming, since well, it was really 2007. So on one hand, that is coming up on 20 years, which is hard to believe, dating myself a little bit. On the other hand, it's like wow, it doesn't exactly feel like yesterday, but it flew by. You know, engineers and even companies that are building platforms, building services, don't really understand how the CDNs work, what the networks really do. You know, and you experience this right when there's a big event, you know something like. You know, netflix live streaming fight. You know some of the comments that are just so uninformed and I think we're way past 20 years into this. How can people still not understand? And so you know, okay, that's fine. You know, I'm not suggesting those companies or those specific engineers don't know anything about video, don't know anything about streaming, but they don't understand how the fundamental, how the roads work. You know the pipes, how the content flows.
Speaker 1:So one of the things that I know you know you've given me a little bit of data about i3dnet and I'm you know this on your website and you presumably talk public about it is the number of direct peering relationships that you have. Talk real quick about why that's important. And I also find something interesting. You did reference POPs. You know points of presence interesting. You did reference pops. You know points of presence, but yet you know you didn't use that as your primary anchoring. You know we have 147 pops around the world, or we have, you know, whatever the number is big, or you know, whereas you go to everyone else's website and that's kind of the very first thing that you see generally, as if that number is bigger, it's better, if it's smaller, if it's smaller, it's not so good. Explain why peering relationships are such a big deal. And you know why 9 000 makes you. Is it the second most um interconnected network in the world? So yeah, explain that for us.
Speaker 2:So pops in general. More is not always better. It really depends. How do you have all these points of presence connected and do you actually have the connections peering connections, for example, with all of the ISPs in that specific region to cover your large population areas directly? So the second aspect is can you connect all of these points of presence together into one big global network, or are you creating small little islands, in which case it's still good for don't get me wrong still good for a large number of application use cases, which is also why a lot of, especially CDN providers put POPs prominently on their websites. But if you have applications which do need to talk to each other between regions, which need to serve traffic between regions, pops is not the primary aspect that people find important, in other words video, video traffic, where it may be asynchronous or or synchronous, I guess two-way, in other words, not asynchronous, um, synchron.
Speaker 1:That's the point. We don't want it to be asynchronous very low latency real time. So, for all of the talk at NAB, where there will be year after year, it's low latency and reducing jitter and all this kind of stuff what you're starting to explain and hit on is actually what people need to be asking their CDN, their networks, not the number of pops. Although it is relevant, it's still relevant.
Speaker 2:Yeah, but let's say, if we let's even look at the recording that we're doing right now Like let's say we had five, six people in here recording all at the same time yeah, then it certainly becomes very important for the central application to be able to reach us all in a good latency and a good throughput and also have the place where the recording takes place be kind of in the middle point of all of us in the call right now.
Speaker 2:And that's actually how a lot of advanced video and voice session application nowadays work. They look at who are all my users in my chat room. They could be friends from each other from different parts of the world and what's the most optimal location for the server to host all of these people together so they all have the most acceptable latency and thus quality for each other. And it works the same in gaming when you decide we have 64 players, we know which regions these players are from, what's the most optimal location to place the server on? And then you're really dependent on the locations being good and connected to all of the other locations in the region.
Speaker 1:And then why does?
Speaker 2:peering matter, because a peering session is, in essence, an agreement with the other party, in this case, most often an isp, of consumer isp, where you say to each other these are my ip addresses, okay, these are mine ip addresses. If we go one, if you have traffic which wants to go from us to you and vice versa, we exchange it, usually for free, if it's not crazy out of balance. In the locations that we interconnect with each other and let's say with the largest isps, of course, you interconnect on many, many different places inside of a country, which sometimes countries, in the case of europe, and that allows you to exchange those bits and bytes truly locally. So if it gets generated local, it will stay local and if there is any incident with the users or vice versa with the application, you then have a direct contact that you can reach out to to get information, contact that you can reach out to to get information.
Speaker 2:So the peering sessions and having so many of them at one point yes, it does help also keep your costs a bit more under control, because what you usually do when there's a lot of traffic involved is that you connect with 100 gig or multiple 100 gig ports towards each other.
Speaker 2:And then you're not paying per gigabyte anymore, no, you just pay the flat fee for the whole port. So it's also very predictable that any traffic going over that port will have a fixed amount of costs. So it becomes a little easier to handle Versus if you, let's say, you don't have those peer-in-connections, you are sending your traffic to a middleman to reach that ISP, and it could be going from middleman A to middleman B to middleman C before it finally reaches that ISP, and then you lose all the guarantees that you have on the throughput, on the packet loss and on the latency of the routing. It gets a little bit technical, but to provide the best possible experience that I would want to have as a gamer or as a, consumer of a voice of video game it is very important.
Speaker 1:Well, yeah, thank you for explaining that, and you know you're going to be in the NetEnt booth. We're very happy to be, you know hosting you. So there's no doubt that you know. I know you're setting up your own meetings and you know you have customers coming to see you. But you know, I think we're going to be able to send a lot of folks your way, you know, to learn how you can help. I want to get your perspective on something you know.
Speaker 1:Cost is right now, you know, the big topic. It doesn't matter if you're selling CDN, like i3dnet, if we're selling VPUs, like NetEnt does, selling software. Almost the very first thing out of a prospective customer client's mouth is hey, cost is our number one driver. How can we reduce cost? What are your costs, et cetera, et cetera. So I'm curious.
Speaker 1:One of the things that I've just observed and heard from the industry is buying cloud services, buying network services, which CDN would clearly fall into that. Sometimes they're sort of hidden costs or you know there's traps in it so it can look on the surface like, oh, wow, this is amazing. This company, you know they're, you know per gig transferred is you know fractions of a penny, you know whatever. I don't even know what the going rates are now, but you know they're getting very, very low. But then you know you check with that same company a year later, two years later, and they're complaining about their costs being out of control. What are some of the? You know the traps that companies can fall into when choosing an infrastructure provider and you know how might i3dnet make it easier for them not to fall into that. You know if they work with you.
Speaker 2:I think that's a very good question and it has a number of different aspects, like we talked about the network optimization questions earlier, the network optimization questions earlier and what.
Speaker 2:You see that with a number of clouds that they do offer a standard network product against an X price per gig, but then you later find out that it doesn't really work that well for your applications, which do need a little bit more of that geographic connectivity, because the default product is that they send all of your traffic out as quickly as they can to the third-party providers.
Speaker 2:So then you suddenly realize you need to buy the more premium networking product which does include all of that regions and the backbone connectivity. And to give you an example from gaming, there's a number of large games out there right now where they also realize that during the evenings about 15 to 25% of the players in a specific region were actually coming from another region to play together with their friends. Or because of time zone differences, it might be an evening earlier, the service might be more full. That certainly becomes then a necessity to buy. When you first didn't even find it on the website, you only found it in the fine print. Later. There's also regarding the kind of solution that you buy, because nowadays you can also buy serverless kind of solutions or SaaS based solutions where you pay, for example, per stream, and that's great when you start up.
Speaker 2:And then when you multiply that by a million, plus when you would- be, successful and you look at the cost, like maybe maybe I need to look at something where I pay per instance or per machine and do a little bit more work myself, because then the costs are a lot more fixed.
Speaker 2:You know how many streams I can output on a machine and especially then if you go with dedicated hardware versus virtualized hardware.
Speaker 2:On the dedicated hardware you can increase the performance and output even more by, for example, utilizing VPUs and you don't have the disadvantages of noisy neighbors who are also trying to claim resources at the same time that you need them.
Speaker 2:So that's definitely a big cost trap, especially when utilizing virtualized instances. That's definitely a big cost trap, especially when utilizing virtualized instances, scaling that up in a flexible way, because we all know cloud is the most flexible option that you can use. But if you have your millions and millions of streams there and still in the most flexible solution, then you're paying too much because how much of that load that you have in the cloud is fixed and doesn't need all this flexibility or doesn't need up to a per minute flexibility. And when you realize that, you're thinking, wait a minute, I could use cloud only maybe for a 10% or 20% of flexible layer that I need on a per-minute-based billing, and then the rest can be on these dedicated machines where I pay per month or per day or per hour, with one of the solutions that we have that we call the flexible metal.
Speaker 1:Yeah, yeah, I like that. I like that name flex metal, Flex metal, so that you pay per hour.
Speaker 2:Yes, it takes some time to provision a machine because it needs to be physically installed. It's not like within two minutes on a VM, sure, but it allows you to bring a larger portion of your daily curve or of your daily peaks. Bring a larger portion to the most, let's say it predictable cost infrastructure, where you gain the optimizations and the cost advantages compared to paying for the ultimate flexibility of the cloud, exactly, and NetEnt VPUs are going in your flex metal solution, correct?
Speaker 1:And then they're also available for a customer that may just need to reserve a certain amount of capacity. Yeah, correct, yeah, this is something you know. Thank you for explaining that, because the notion of a hybrid cloud has been talked about for years. Unfortunately, it hasn't really been realized, especially when it comes to hardware. For some very practical reasons, if you're going to bring an accelerator in or any sort of dedicated hardware, so like a VPU, that A needs to be available in both locations. So it needs to be in your dedicated data center, your on-prem or you know wherever the customer's operating, call it their primary, you know, core service, but then it needs to be mirrored or reflected, you know, in the cloud you're flexing into. And, by the way, this is true of GPU, this is true of any other technology, so even GPU, although they're much more accessible, extremely accessible in all the public clouds, maybe the particular version of GPU somebody's using is not available. So then it forces people to have scratch their head and either compromise. Either compromise, maybe on quality, because they have to use a GPU that's not, it doesn't have as good of an encoder, you know, or they're just paying way too much because all they have is, you know, something that's way overpowered, but you know it can handle the video. So we're really excited about what we're doing with you and what i3dnet is doing to deploy VPUs, because the market has been asking us for this for well since we launched. So it's something, and it's taken a little while for the ecosystem to develop, but we're now at a stage where it's progressing really, really rapidly.
Speaker 1:So I think I'd like to wrap up our conversation here at this point, because what I want to do is implore our listeners if you are listening before NAB and if you're going to be at NAB, make sure that you get a meeting with, you know, with Stefan, with the rest of the team that's going to be there. You know with Stefan, with the rest of the team that's going to be there, the i3dnet team. And you know, if you're running between meetings and you need to sort of just swing by, they're going to be in the NetEnt booth. They'll be in the NetEnt booth, the whole show. So they you know they have a kiosk, there's a TV set up there. They're going to be, you know, holding meetings, doing demos, talking, you know, talking to anybody who wants to learn more. So definitely you should stop by and see them. So, stefan, you know parting. You know parting words of insight or wisdom parting words of insight or wisdom.
Speaker 2:Yes, if I have to give one big piece of advice is, when selecting your infrastructure provider, don't just look at the initial nice things that you might get, like, oh, you get a few hundred K of credits and they want you to really build everything on it and it's so convenient and flexible.
Speaker 2:Now keep in mind in the back of your head, how will my costs and product look when I am successful and scaling as a champion?
Speaker 2:Yes, and then you start to realize okay, maybe I should not go with something that risks me a very big login or where I ultimately am stuck with a massive million spent commit to get any discounts. Maybe I should go with a solution where I combine, maybe, different providers so I have negotiation strength with all of them. So it's not just all about, maybe, the brand recognition, of going with the most common names out there. It's all about having the right architecture with the right cost model, the right network capabilities with the correct trade-offs that your application needs Because, again, not all applications are the same. Some might need less concern on latency but more on throughput, or more on packet loss, or more about squeezing as much as encoding as you can get out of every single physical machine, but you should always focus and think very hard. What would it look like when my application is successful and ensure you are with a partner or streaming platform that is ready to scale, then with you without any nasty surprises?
Speaker 1:Yeah, that's great, that's awesome. I totally agree, so excellent. Well, thank you again for joining us on Voices of Video. I really do look forward to in about a month just a little bit longer than that being face-to-face and spending time with you and the team. So, for all of our listeners, please do schedule a meeting with i3dnet. Stop by our booth, stop by NetEnt. Love to talk with you and, in the meantime, happy encoding, happy streaming. Have a great weekend and we'll see you all soon.
Speaker 2:Thank you so much, Mark, and all the listeners.
Speaker 1:This episode of Voices of Video is brought to you by NetInt Technologies If you are looking for cutting-edge video encoding solutions.
Speaker 2:check out NetInt's products at netintcom.
