KubeCon + CloudNativeCon EU 2026 | Bill Mulligan, Isovalent

Clips
More from KubeCon + CloudNativeCon EU 2026

Bill Mulligan

Cilium and ebpf.io Maintainer

Cilium

play_circle_outline Cilium as a Platform: Gateway API, Load Balancing, Hardware-Aware Networking, and Migration Guidance

play_circle_outline Identity-Based Networking for Transient AI Agents, VMs, and Containers: Security and Observability at Scale

play_circle_outline Goal: become the standard networking layer across clouds and workloads; boring equals reliability.

play_circle_outline Cilium 1.19 highlight: native mTLS via integrated Ztunnel from Istio Ambient mode.

Info
Transcript

Bill Mulligan, Isovalent

Bill Mulligan

Cilium and ebpf.io Maintainer Cilium

In this interview from KubeCon + CloudNativeCon EU 2026 in Amsterdam, Bill Mulligan, Cilium and eBPF maintainer at Isovalent, joins theCUBE's Rebecca Knight and Rob Strechay to discuss how Cilium is evolving from a container networking project into the foundational networking layer for VMs, containers and AI agents. Mulligan reflects on Cilium's 10th anniversary, tracing its origins from solving identity-based networking for early container environments to becoming a comprehensive platform spanning load balancing, API gateways, ingress and security. He explai... Read more

explore Keep Exploring

What are Cilium’s major capabilities and current priorities, and what should organizations consider when migrating away from NGINX Ingress (especially regarding the move to Gateway API)? add

How is Cilium adapting to the challenges posed by AI workloads—such as more agents, increased service-to-service communication, and more dynamic and unpredictable traffic patterns—and what does the future look like for its role in identity-based networking? add

How do you expect Cilium to evolve and what role will it play in networking and infrastructure over the next decade? add

What are the key highlights of Cilium 1.19 that users should be aware of, and what implications do they have going forward? add

bolt Powered by CUBE AI

Bill Mulligan, Isovalent

search

Rebecca Knight

>> Hi, everyone, and welcome back to theCUBE's live coverage of KubeCon + CloudNativeCon here in Amsterdam. I'm your host, Rebecca Knight, alongside Rob Strechay. We're talking about boring, being a signifier of success.

Rob Strechay

>> Yes, I think that it's funny that it's been called boring because I think this is one of the hardest topics, especially in this agentic AI realm that we're embarking on with Kubernetes right now.

Rebecca Knight

>> I am all in favor of making anything exceedingly complex, boring and simple. So, with that, I would like to welcome Bill Mulligan, Cilium and eBPF maintainer at Isovalent onto the show wearing the uniform. I love it.

Bill Mulligan

>> Of course.

Rebecca Knight

>> Looking great. Looking great. So, Bill, Cilium is celebrating its 10th birthday this year, and this is very exciting. Before we get to where the project is heading, why don't you take us back to the very beginning about its founding and what it was all about and what problem it was trying to solve?

Bill Mulligan

>> Yeah, definitely. Like any good idea or bad idea, it started with a small group of friends, probably over some beers at some point. And I think one of the things that I like about Cilium is you can go back and to git history. It was open source from the first commit. It was like license file, read me, and that was it. And since then, it's all been in the open. A lot of times people are like, "Oh, we need to think about these ideas." No, it's like, "We're going to build in the open. We're going to create this community around the project." And if we zoom back 10 years ago, that was just the beginning of containers. There was the whole SDN movement, it was happening, but we're looking at, "Okay, what happens when things change a lot more quickly? How do we deal with the churn of identities?" And that was the problem that Cilium originally set out to solve. How can we do identity-based networking and how do we solve that at-scale?

Rob Strechay

>> Yeah. I mean, to me, Cilium has so much in it. I mean, it's not just one thing. It's almost, I mean, not to use the P word, but it's like a platform withinside it.

Rebecca Knight

>> The P word.

Rob Strechay

>> Yeah, but when you start to look at it from load balancing to API gateways, ingress and egress proxies, helping from a bandwidth perspective. When you start to look at this, I mean, you have so many large companies using it. What are the requests that are coming in at this point? I mean, I hear the boring word and I'm like, okay, well, there's still got to be, "Hey, here's where we want it to go now, because this is what's changing," or, "these are some of the challenges we see."

Bill Mulligan

>> Yeah, definitely. I think the hard and also exciting thing about Cilium is it does cover so many different areas. And when I look at the developments happening in the project, it's like, "Whoa, we do all of this together." So, just two examples that are really big in the community right now. There's the whole deprecation of NGINX Ingress and Cilium is coming in as both Cilium ingress and Cilium Gateway API to be able to allow people to get traffic into their cluster, routed across that. And then, if you go all the way from layer seven and zoom down to the hardware, there's the whole like, how do we integrate the hardware into Kubernetes? How do we make Kubernetes hardware aware for things like AI or really big networking workloads, so things like DRA? And Cilium is integrating DRA into that. How do we have multiple network interfaces on a single pod? And, so it's really like this whole stack of networking all in this one project.

Rob Strechay

>> You brought up the deprecation or actually just the end of life for the ingress... Easy for you to say. When I look at that and there's a lot of things that slightly overlap, right? You got Envoy and you have some other things that are out there that could theoretically be used. What should people be thinking about as they're moving away from that, given that we're less than a month away from the end of life and people have to have these migration plans?

Bill Mulligan

>> Yeah, definitely. I would think about, "Okay, what are my most immediate pressing concerns and like what are the features that I'm going to do?" People sometimes are like, "Okay, give me a whole checklist of everything it can do." And it's like, "Well, what are you actually trying to solve?" It's the classic question, like, "I need a service mesh." It's like, "Okay, for what? Is it the networking part? Is it the observability part? Is it the security part?" So, I would one, map out what are your actual requirements? What are you using it for right now? And then, look a little bit about, okay, what features are we going to need in the next six months, year? And so you can come like, these are our requirements rather than being like, "Oh, this is like the best thing." And then, I would seriously consider, yes, Cilium does have an ingress, but ingress is not the future. All development is going into Gateway API. One of my colleagues is actually giving a talk and I think he has like 2,000 attendees signed up on shed right now for that. So, the interest, the movement is going into Gateway API. So, yes, you could go to Ingress, but everybody's moving to Gateway API. So, that's probably what I would recommend if you can.

Rebecca Knight

>> So, to that point though, but AI workloads are changing what infrastructure needs to do at a pretty fundamental level. More agents, more service-to-service communication, more dynamic and unpredictable travel patterns. How is Cilium holding up against all those increased demands and what do you see the future holding?

Bill Mulligan

>> Yeah, definitely. So, actually going back to what I said at the beginning, we're talking about how do we do identity-based networking at scale, right? That's what Cilium was founded on and now 10 years later, I can see the exact same thing. What do you have to worry about with AI? Well, you have to worry about identity. You have all these new agents in your infrastructure. You're doing it at an unprecedented scale and it's all about the network. So, that's what Cilium's trying to solve. I think the core of what Cilium doing is giving it an identity in the network. It's making everything a device on the network. And before it was containers, but now we're going into a world with agents, giving them an identity, being able to control them, being able to do security and observability based on those identities, and then even rewinding to other things like things like, okay, now people are also bringing VMs into the Kubernetes and cloud-native world. Well, guess what? That's just another device on the network. So, you can do containers, you can do VMs, you can do AI agents, giving everything identity, putting it on the network and being able to do networking, observability and security around that.

Rebecca Knight

>> And it really gets back to the idea of making boring being the differentiating factor in an industry that's obsessed with what's new and exciting.

Bill Mulligan

>> Let's go. Let's go. Let's go.

Rebecca Knight

>> Actually, it's reliability and dependability that is the hardest to pull off.

Bill Mulligan

>> Yeah, definitely.

Rob Strechay

>> And it would seem that a lot of the users in the community that are using Cilium, I mean, again, it's been around almost as long as Kubernetes has now. I mean, it's off by like a year and a half there, something like that. And when you look at this and you look at those three things, the networking, observability and the security, to me, there's so many different pieces, from an AI perspective, that go into each of those, right? Because you look at AI and it's like, to your point about agents, it's like, who's it talking to from where to what? Okay, I need to make sure that it can't get there. It doesn't want Robin randomly in Japan for some reason logging in with Rob's identity, but he shouldn't be able to get to certain resources from certain places. Or network transparency and being able to have the security done where it's being able to route the things the right way so that traffic only has that one pattern to go through. Is that a lot of the pressure you're starting to see across that stack where it's a use case that is touching networking and observability and the security aspect of it?

Bill Mulligan

>> Yeah, it's definitely a lot of that. And I think what sets Cilium apart is doing everything based on eBPF because when you think about all these things, like we have a lot of software running and it's all running on hardware. And what's the thing interfacing between the software and the hardware? Well, that's the kernel. Everything is running on Linux these days, everything goes through there and eBPF can see everything that's happening. And so, by using eBPF as the core of Cilium, we're able to integrate all the signals across networking, observability and security, correlate them together, take the context that we know from Kubernetes, from the software, and then, okay, should this application be able to access these hardware resources? And you have really clear insight into that and everything has to go through the kernel.

Rob Strechay

>> And I mean, you kind of mentioned it, but just to double click on it, when you look at, because this is the whole challenge and opportunity, I would say, with agents is agents are basically going to act like another person logging in and doing some set of activities for a certain time and then go away and come back and have different sets and things. Are you seeing that that transient nature, which has been typical with the ephemeral nature of containers and where it came from, has that been, "Hey, this isn't as hard as we thought because we have this heritage or..."

Bill Mulligan

>> Yeah, it's the same challenges, looking slightly different. I think that's the cool thing about being in the kernel, at the end of the day, AI is just another workload. And being able to be in the kernel, you can see everything that's going on. And then, I think what's different now is we need to translate instead of saying, originally it was like, "What is this VM doing?" Then it was like, "What is this container doing?" And now it's, "What is this AI agent doing?" And so, what you need to do is take that kernel context that is just like, okay, this process tried to open this network connection, this process tried to access this file, and you need to apply the context that you know, is this a VM? Is it a container? Is it an AI agent? And that's like what the abstraction of Cilium does. That's why it's become such a powerful platform because it takes this really low, granular, but super raw data and correlates that with the context that we have to give you the abstraction that you need. So, you're not just like dug down into the eBPF byte code. Nobody wants to write eBPF programs. Nobody wants to hire a whole kernel team, but Cilium gives you that for free.

Rob Strechay

>> Yeah.

Rebecca Knight

>> Something you said struck me as so profound, which is at the end of the day, AI is just another workload, which we talk about on theCUBE that AI is revolutionizing work in itself and changing everything, but that's really true, what you just pointed out. How do you think about the next decade of Cilium in terms of the bets you're making today, the things that you are hoping will work out and what keeps you up at night?

Bill Mulligan

>> Yeah. So, the next decade of Cilium I think is actually, I know this isn't exciting or sexy, but be more of the same. I think like first we saw Linux become the operating system, then we saw Kubernetes become the operating system of the cloud. And what we're really seeing Cilium become is like the networking layer. Working at Cisco, Cisco really defined the modern data center, it took things like TCP/IP, BGP, burned it into silicon, you put your top rack, switch it in and there was this whole generation of people that like, "I built my career on Cisco switches." But I think what we're going to see in this next decade is people are going to say, "I built my career on Cilium because it's how the network runs for VMs, for containers, for AI agents." And that's really powerful. Once something becomes a standard, it's what everybody knows and operates. So, I think what we see a lot of people are turning to Cilium because it's like nobody's running in a single cloud, nobody runs a single type of workload, but what Cilium allows you to do is have a same consistent networking experience that you know how it works across any cloud, across any type of workload. And it gives you a way to build your infrastructure, build your career, build your company in a way that is boring. And that's what people want. No one wants to be pulling out switches at like 2:00 AM and now no one wants to be like on call getting the page at 2:00 AM. The same thing, people want infrastructure that works and that's what Cilium gives you across all these disparate workloads and compute.

Rob Strechay

>> So, as somebody who was a CCNA and learned routing from Radia Perlman, who rest her soul there, but from a perspective and had gone down and done BGP and all of that, I have the scars to... I totally agree with you. Boring is awesome and not getting into the code all the way and not having to know your AS numbers and get all of this other stuff, which was crazy. But one of the things you also had is that Celium's release of 1.19 just came out, right? So, what were some of the highlights of that that people should be aware of and as you look forward here?

Bill Mulligan

>> Yeah, definitely. I would say the biggest one that people are most excited about is around MTLS. So, Cilium took Ztunnel from Istio Ambient mode and integrated that in. And so, it gives you native MTLS directly into Cilium. So, this is one more piece in the stack of things that Cilium provides. And I think people will be excited about this because there's a lot of complexity around service mesh and now it's like, okay, Cilium is already as installed as your C&I. If you have regulatory requirements or compliance or anything else and they're like, we need pod to pod MTLS. You can be like, "Okay, one more flag we need to turn on," and done, that's it.

Rob Strechay

>> Yeah. No, I would agree. I would agree. I think to me that is one of the big things out of it that I saw. So, good to hear.

Rebecca Knight

>> Well, Bill, a fantastic conversation. Thank you so much for coming on the show.

Bill Mulligan

>> Yeah, thanks for having me.

Rob Strechay

>> Yeah, thanks.

Rebecca Knight

>> Be back in your neck of the woods in a couple of years.

Bill Mulligan

>> Yeah, I can't wait for Berlin.

Rob Strechay

>> Yeah, that'll be a lot of fun.

Rebecca Knight

>> In 2028, I believe. Yeah, excellent. Excellent. Stay tuned for more of theCUBE's live coverage of KubeCon + Cloud NativeCon 2026 live in Amsterdam. I'm your host, Rebecca Knight, alongside Rob Strechay. You're watching theCUBE, the leader in enterprise tech news and analysis.