Voices of Video

Your Buffering Wheel Is Not a Feature: Why Real-Time Video Lives at the Edge

• NETINT Technologies • Season 3 • Episode 41

Share episode

Real conversation dies the moment latency enters the room.

In this episode of Voices of Video, we break down what truly separates traditional streaming from interactive streaming and why the old playbook of centralized encoding, deep buffers, and best-effort delivery simply cannot support audiences that talk back, transact, and co-create in real time.

We start by defining the hard technical requirements of interactivity: ultra-low latency, low jitter, and deterministic paths from first-mile compute to last-mile delivery. From there, we explore why pushing compute closer to users is necessary, but not sufficient. Owning the backbone matters just as much as owning the servers.

You’ll hear how private long-haul circuits, strategic peering, and unique subsea routes, like a direct Fortaleza-to-Portugal path that avoids U.S. detours, can shave 50–60 milliseconds off round-trip latency, transforming global collaboration into something that actually feels local. That performance shift unlocks new many-to-many use cases across gaming, telehealth, webinars, watch parties, and live commerce, where interactivity directly drives revenue and retention.

We also get practical about architecture. Flexible bare metal with bandwidth-rich instance types, API-driven provisioning, and Terraform automation allow teams to scale capacity in minutes when demand spikes. On the network side, we explain why peering, cross-connects, and router upgrades require lead time, and how a pre-built baseline protects quality under load.

Finally, we zoom out to strategy: hybrid models that combine a dedicated edge layer with elastic public cloud, integrated through open protocols that avoid vendor lock-in. The goal is simple: give builders control, keep users close, and make real-time video feel effortless.

If you care about real-time video that actually feels real, this conversation is for you.

Topics Covered

• Linear streaming vs interactive streaming
• Why edge compute beats centralized encoding for real time
• Many-to-many use cases across gaming, meetings, and telehealth
• Backbone ownership vs CDN dependency and third-party transit
• Subsea routing strategies that cut 50–60 ms of latency
• Bandwidth-rich instance types and API-based provisioning
• Scaling network capacity ahead of demand
• Hybrid architectures combining metal and public cloud
• Open protocols that avoid vendor lock-in and build trust

Links & Resources

• Voices of Video podcast
https://netint.com/podcast

• i3D.net – Global edge and backbone infrastructure
https://www.i3d.net

• NETINT Technologies – Video encoding ASICs and platforms
https://netint.com

• Learn more about edge video architectures
https://netint.com/resources

This episode of Voices of Video is brought to you by NETINT Technologies.
If you’re building high-performance, power-efficient video infrastructure, learn more about NETINT’s ASIC-based encoding solutions at https://netint.com

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. Voices of video. Voices of video. Voices of video.

Wouter Koppert, i3D.net:

Yes, thank you very much, everyone. Very nice to be here. Second session for me. For those who don't know me, my name is Malte Koppert, a business element manager within ITANET infrastructure as a service provider. Based in the Netherlands, and we have more than 20 years of experience in mainly uh hosting the low latency multiplayer games across the globe. And right now we're also investigating and tapping into the market of streaming. And in this talk, I've invited my colleague Martijn Smid, which is our VP of network at i3d.net. And we're gonna dive deep into the differences of streaming and interactive streaming. So Martijn, welcome. Thank you very much. If you're ready, we can uh we can start off.

Martijn Schmidt, i3D.net:

Yeah?

Wouter Koppert, i3D.net:

First question the big question. Martijn, what is the big difference between uh streaming and interactive streaming?

Martijn Schmidt, i3D.net:

Thank you, Walter, and thank you for uh having uh us. So uh the difference between the interactive side of streaming and the more traditional linear side of streaming is that uh you can tolerate a lot more things like latency or buffering uh on the uh on the stream, right? So when you want to actually interact with your audience in a in a live manner and you want to get two-way feedback rather than one-way feedback, you cannot afford to have like a very big amount of jitter or a long buffering delay to avoid any uh performance issues. So uh when it comes to that topic, uh you need to have more uh compute near your edge rather than having uh centralized compute where you're doing one time the uh processing of the video uh towards uh the format that you're gonna be viewing it in. And uh yeah, that's uh where we believe that with the uh i3D.net infrastructure we can bring that compute at scale at over 60 different locations in the world.

Wouter Koppert, i3D.net:

So okay, Martin, thanks. Very clear answer. Uh second point that I want to tap into basically is there are different streaming models, right? As in uh one-to-many or many to many. Can you tell a little bit about how our network uh is doing in the different styles of these streaming modules?

Martijn Schmidt, i3D.net:

Yeah, so on the side of the many to uh or one-to-many, uh, I think that's not really our niche. I mean, we can do it, but uh, there are many other parties out there who can do this type of traditional streaming. Uh, where I think that we shine is like the uh many to or one-to-few or uh well many-to-many type of streaming, uh, which is where you really have more interactivity. Uh, you have like uh interaction with the audience and can take the form of like a video call or like a webinar or like a meeting where you are uh presenting something to an audience, but it can also be as simple as, for example, um doctor's visit, online doctors' visit, so telehealth. So a lot of different uh type of uh things that exist out there when you think beyond the simple uh yeah, for example, sports matches.

Wouter Koppert, i3D.net:

Uh I was talking about the delivery part basically. So can you tell us a bit more about how we compare to a regular CDN services, for example?

Martijn Schmidt, i3D.net:

Yeah, so where CDN services have really placed a lot of different uh hardware usually into the network of the ISP itself, and they often do not operate their own backbone, uh so they rely on basically third-party connectivity to get between the different uh locations with all the implications in terms of quality that that may bring. Um, we actually own our own compute uh in the data centers where we are the tenant, right? So we are not dependent on somebody else's goodwill to be able to make hardware changes if that's needed. Um, so for example, we can install the the net in VPUs that we are uh looking at here today. Um, but also we have built our own backbone, uh we own our own network. So uh what that means is that when you want to connect, for example, from Warsaw to Los Angeles, um, we can carry that all the way on our own backbone that we have built with sub-C cables, uh long distance circuits, and uh so on. And in fact, even if you are uh connecting into our network uh, let's say to a server in Los Angeles, you can still, as an end user in Poland, connect all the way into that uh other location on the other side of the world. And uh that doesn't only bring benefits in terms of stability and predictability, um, but when it comes to uh specific routes, for example, from South America to Europe, we also have our own sub-seed cable on unique routes like uh ALA Link, which connects directly from Fortaleza to uh Cines in Portugal. So by that way, you are basically able to literally cut the corner uh by not uh connecting through the United States on the way to LOTAM, but actually going there directly, it can easily save you like 50 to 60 milliseconds. So it's quite a big difference uh in terms of the uh the performance when it comes to real-time content.

Wouter Koppert, i3D.net:

So I want to zoom I want to zoom a bit more into the workloads as well, right? Because um obviously it's uh it's a bandwidth consuming workload that's running on that uh under the infrastructure, so to say. Can you zoom in a bit more on what type of workloads require an interactive environment?

Martijn Schmidt, i3D.net:

Yeah, so for example, uh we have uh quite a few customers who are doing uh video streaming for uh meetings, for example, or for group chats, right? So they might connect together with their friends online while playing a game, or they might uh connect to have like a visit with the doctor, uh so on and so forth, right? Um, and it's really that uh interactive type of streaming where we can make uh the difference, and then depending on the application that you're gonna be running, uh you might want to optimize for having more memory, or you might want to optimize for having more uh bandwidth in your uh in your machine, and we can make those changes. Uh and in general, we have like several uh instance types, uh, and those instance types you can order with an API uh on demand and scale up as it's needed uh to support your actual workload. So, for example, if you have like a busy day and uh you didn't prepare for it, you didn't plan for it, you can still, without even talking to a human, uh, get your extra capacity up and running through our flex metal service, where uh using Terraform uh you can uh yeah uh basically set up the instance type that you want to have, and uh the system takes care of the rest, and a couple of minutes later you get the login credentials to your machine and you can install your applications.

Wouter Koppert, i3D.net:

Alright. I want to also talk a bit more about the networking part because now you zoomed in on the computer, the hardware itself, and you have to handle uh live streams, for example. Scalability is an important factor. Can you tell us a bit more like how our network is scaling with the demand or live stream, so to say.

Martijn Schmidt, i3D.net:

So the difficult part when it comes to uh network is that uh it can take quite a bit of time to actually scale up. And uh the reason for that is that you need to negotiate with the other ISPs about having upgrades in pairing capacity. You need to have like uh routers physically already in place, you need to have the data center operator running physical cables between the equipment of you and that of the other ISP that you negotiated with. So there's quite a bit of lead time uh between actually deciding that yes, you want to have that upgrade up and running and actually having it in place. So by the time that you start thinking about it when you actually do like a live event, it's usually already too late. So that's why it's important to already have like a solid baseline of uh infrastructure in place that you don't need to start building when the customer comes, or that you don't need to start thinking about how to interconnect with another ISP when that first order comes in. And uh, in that sense, our decades of experience means that we have built that base layer, we have that infrastructure, uh, we've been doing this for a very long time. So, as a direct result of that, you get a reliable local connection to your end users uh without having to go out and build it yourself with all the delays that that entails.

Wouter Koppert, i3D.net:

But then where did you see the future of networking going towards soon when you especially look inside the world of streaming?

Martijn Schmidt, i3D.net:

So, my expectation is that on the one hand you will keep having the traditional type of streaming where uh people are just all interested in the same content. So, on the one hand, you have like the traditional type of streaming where uh everybody is looking at exactly the same content. Um, but as time goes on, and now that the internet is actually getting to a point where it can support this kind of interactivity at scale easier, it's also a matter of time until people start finding new ways to use streaming to uh generate content and business, right? So, for example, one of our customers was um using uh streaming to a smaller audience and then having interactive orders basically that they could give people, right? So rather than giving a twitch shot, for example, um you would be saying, okay, um if you do this and this action or you achieve that and that kind of thing, like you need to do this amount of damage, or you need to have that many kills in the game, or whatever, then you can get the reward and otherwise not. So you have a much more direct interaction between the streamer and the audience, and uh that can, especially for the audience, bring like a lot more feeling of closeness, right? And by that audience engagement, we think that the streamer can like earn much better revenue than they would be able to if it was just one-way linear, and at most, you can react on what's happening in the chat.

Wouter Koppert, i3D.net:

Martin, what I also want to discuss with you is something related to a topic that we see as obviously our background is in the gaming industry as well. Um, but we saw over the years that there was a big migration towards the cloud providers for the easiness of use and the flexibility that it has. Um, what what we see very well right now is that people more often choose for some sort of a hybrid environment where they bring their base layer into an infrastructure as a service provider like us and they use the scalability for the cloud. Do you think that there is when you look at icd.net, for example, with the new products that we have, the build of the flux iPhone product, that there is more opportunity for us to scale more into the upgrowing demand in the live streaming uh industry.

Martijn Schmidt, i3D.net:

Yeah, I think what is really the hallmark of the cloud providers, it's like easy to get the infrastructure, right? But when you're actually in there, it's very hard to move away and to move out. Like they are not designed to play nice with each other. So what is making us different is that we are more open to like integrating at different layers of your company, right? So you have, for example, customers that might want to have like metal in one location with a specific uh type of hardware configuration in another location, and then at the same time, they might not even need that because they have their own data center, their own network, and so we can deliver a service like IP Transit to them to support their business in that sense, right? So it doesn't really matter at what level or at what layer you want to integrate your business, uh, you can do so freely and on the basis of open protocols. So we will never go to a situation where we lock you into using our proprietary protocol. At the end of the day, I think the best way for a customer to feel bound to the company is as if they are there voluntarily, they know that they can leave. That's how you can like create the trust that okay, hey, you you have that social contact with the customer that you need to keep doing good, otherwise they can literally pack up and leave, right? So that's I think a very healthy type of thing where we are performing in a different way than uh cloud providers.

Wouter Koppert, i3D.net:

Okay, what I know. Thank you very much. All my questions are are are answered basically. Is there anything that you still still want to add? Any last words for for the audience?

Martijn Schmidt, i3D.net:

Well, I think I'm uh I'm good. Um are there any uh questions from the audience? Or uh all right. Thank you very much then and uh enjoy the rest of the event.

Voices of Video:

This episode of Voices of Video is brought to you by Netting Technologies. If you are looking for cutting edge video encoding solutions, check out netting products at netting.com.

Head Shot

Mark Donnigan

Host
Head Shot

Anita Fejter

Producer