Cloud Networking: Embracing the Chaos

Daniel Newman:
Hey, everyone. Welcome back to The Six Five Summit. Daniel Newman here, CEO of the Futurum Group. Another great conversation coming your way. We’re in the cloud infrastructure track. We’re going to be talking to someone that needs no introduction here at the Six Five Summit. He’s been an opening keynote with us before. He’s someone that I can call a personal friend, who’s doing really important work in this space, Doug Merritt, CEO, Aviatrix. Doug, welcome back to the Six Five Summit, so good to have you again.

Doug Merritt:
Thank you, Daniel. So good to be here. I wish that we were in person, but it’s still good to do it virtually.

Daniel Newman:
Yeah, you know what, we will get there maybe at some point.

Doug Merritt:
You know what, we will.

Daniel Newman:
We have so many great speakers and thinkers. You and I live in the same town, and just trying to get together to have lunch takes three months. Just imagine trying to get 50 of your peers, CEOs and top executives, get them all in the same place at one time. I think they do it once a year, it’s called the World Economic Forum. I think that’s the only time I’ve seen all people in the same place at one time.

Doug Merritt:
You’re right.

Daniel Newman:
It’s a dream, but we will get there. But these conversations are so important, and there’s so much going on. Hey, I want to talk about the cloud networking and the chaos that’s going on in this space. But before I do that, because Aviatrix, you’re running a new company. Not everyone here at the Six Five Summit has seen you in this capacity. They saw you when you were running another very exciting growth company. Give them the quick one minute, Aviatrix, what are you guys focused on?

Doug Merritt:
Yes, would love to. Yeah, I’ve been at Aviatrix a little over 10 months now. I thought I was going to retire and then realized that I-

Daniel Newman:
Look how young you are.

Doug Merritt:
I’m a bad retired person. I missed the team. I miss being part of a team. But Aviatrix is about a 10-year-old company, and we take the concepts of software-defined networking, which have been around since 2008, ’09, ’10, but we have revitalized them in a completely different way so that we can provide networking services. We can provide integrated networking services, which done properly and we could talk about that, wouldn’t separate out all these different functions, load balancing and firewalling and security. It’d be one beautiful integrated networking capability, the way that networking should have been, but we provide that in a native way inside of all the different clouds out there, but also to all the edge providers. There’s a lot of networking gear sitting at Equinix and Megaport and Digital Realty and other edge providers, and there’s a ton of networking gear within last gen proprietary data centers. So we bridge the gap between that hybrid world and provide a single networking abstraction layer and core transport layer, a whole data plane, that makes networking controllable, seamless, high visibility, a ton of analytics, highly secure across that landscape.

Daniel Newman:
Yeah, and of course, from what I understand and from our conversations and what’s gotten me pretty excited about what you’re doing there is there’s a pretty compelling economic case as well, that you can help people do all those things and at the same time drive some economic value. So you started alluding to it, but talk a little bit about that pivot because it’s kind of legacy versus now the cloud chapter.

Doug Merritt:
Yeah, so if we go back to Cisco and the boom of networking, 1990s and 2000s in the classic data center world, it was… For a while, we all now know because of Nvidia, that Cisco was the most viable company in the world for a while, a lot like Nvidia’s getting there today. Because connecting things was so mission-critical as we’re going to the internet. To be able to serve complex enterprises, the way that networking landscape was formed is a whole series of purpose-built boxes for the different networking functions. In networking, you have to move packets, there’s physics involved, which means to do it properly in that world, you need a dedicated silicon and purpose-built capabilities. So you had your routing boxes, your switching boxes, your load balancing boxes, your DNS boxes, your VPN boxes, your firewall boxes, boxes everywhere. And they made up 10 to 20, even 25% of the typical bill of a data center because it took money to actually build and ship those things.

It created a lot of capabilities and control. It had a bunch of negatives aspects as well. There’s a lot of overlapping capabilities between those boxes that should have been seamless. You had different UIs, different command lines that you had to memorize for something that should have been nice and natural and integrated, one whole thing. Software-defined networking came about in 2007, ’08, ’09. Three professors began to evangelize that. It formed into a company called Nicera, eventually. But their concept was, why don’t we do networking right? Hardware has really progressed. We can get general purpose hardware now and have a combined networking platform. That movement began to take some shape before cloud really intruded and changed the entire world. But then they came forth a whole different networking paradigm.

But that historical context, I think, is important for everybody to understand because we’ve gone through a lot over 30, 40 years of the old seven layer OSI stack and really thinking through what’s ideal in layer one all the way up to layer seven. We’ve gone through a lot of iterations to try and bring greater power and integration and flexibility across those layers and between those layers. And networking is mission-critical in that layer three, layer four component of that stack.

Daniel Newman:
And we’ve also hit a moment with this AI inflection that’s only accelerated the need for networking. The amount of data we’re moving and pushing through the networks and through these data centers is going to be exponential. We hear a lot about compute. Everybody’s focused a lot on computer right now, but we have to store the data, and we have to move the data, we have to secure the data, we have to make it accessible, usable, and we have to keep this stuff… we have to manage policies and stuff. These data, like you said, it’s in the edge, it’s in the cloud, it’s on-prem data centers. There are many different hardware architectures. There’s many different storage architectures that are going on at the same time. There’s lots of different app layers. The cloud was supposed to fix all this, Doug, but what I’m hearing you allude to is, like we always joke about, it’s just in someone else’s data center. The cloud didn’t really fix all this, did it?

Doug Merritt:
No, they didn’t. They fixed a lot of things. So the amazing amount of infrastructure that’s now available at, I think, a different price point than it was if you had to build it yourself is awesome and has powered a lot of innovation around the world. But ultimately, the cloud providers all created a series of services from upper layer services, application and data services all the way down to things like networking services. That turned out to be proprietary to their clouds, which made sense because they understood the virtualization layers, the core infrastructure capabilities, they were trying to optimize what they’re doing within their clouds. For networking, for some odd reason, not totally odd, if you paid attention to the workloads that were coming onto your cloud, the big movement that we all saw 2017, ’18, ’19 and ’20 was let me lift and shift. Let me take my last gen applications that had… And if we go back to networking land, these dedicated networking services that were virtualized software boxes and their underlying hardware components, let me take all that stuff and move it over to the cloud.

And that, as long with modernization, that was one of the big, big, big movements still going on today. The cloud service providers looked up at those networking services and realize, “Hey, we need to provide some of those advanced services as well. We shouldn’t just be a core connectivity framework, so let’s form teams to do that and let’s form a routing team. Let’s form a load balancing team. Let’s form a firewalling team.” And they recreated that box metaphor that STM was trying to obliterate. That creates a whole series of problems. The way that they defined the boxes was different for every cloud service provider, every hyperscaler. They took their own view on what those boxes should be so you no longer had a consistent view of a load balancer or router across these different clouds. There are obvious proprietary set cloud, is typical for AWS to build a set of networking services are pretty low in the stack and say these are going to run better on Microsoft, on Azure than they do at AWS, or as good as.

So you’ve got a whole series of proprietary services that are defined differently within every single cloud that provide some pretty non-desirable lock-in within the cloud. If you can’t get portability at the networking layer, you’re really pretty stuck with what you can quickly and easily shift. You’re not using the CSPs the way that they initially crafted themselves, which is use us for workloads that make sense and shift that around based on your business needs. With LLMs, that’s just going to end in hyperdrive, so. It is inefficient architecture and as you’re alluding to economic chaos and there’s operational overhead, there’s security and vulnerability issues that people are dealing with.

There’s latency issues. There’s a bunch of things that if they could do it over, I bet you they would, that are soft costs, but then there are hard costs associated with that as well. Literal things I pay for in my monthly CSP bill, they appropriately had to create data processing and egress fees. It costs them money to move stuff around those clouds and they’ve got to pay for that infrastructure. But if you are not logical or if you haven’t really thought holistically on how to architect some of these things, then you wind up incurring a whole bunch of unnecessary data processing egress costs, inside of clouds and between clouds obviously for sure. So it turns out that not only is there huge operational efficiency gains and visibility and security and a whole bunch of holistic elements that something like Aviatrix can offer, but that we actually offer in a pretty unique way because we’re the only data plane that lives within the clouds that actually moves packets around. But there are some significant hard dollar costs that we can provide, hard dollar cost savings we can provide to customers as well.

Daniel Newman:
Yeah, so let me do a double click there maybe because you said a lot and you said it efficiently, and clearly, 10 months in, you’ve got it down, Doug. But the bottom line here is that the clouds were pretty well-intentioned in what they were doing in terms of delivering the services comprehensively. A couple of things that most of the clouds didn’t consider for instance, was how much hybrid there would be and also how much multi there would be. And obviously, none of them architected their environments or what AI was going to do. All these things though have had substantial impacts on networking and as data is exponential, you are seeing more pressure on the network. We’re bringing more costs back to move data around. You’re making companies compromise on where data sits because they’re thinking about that cost. And if you’re compromising on where data sits, you might also be compromising on how much data is accessible for workloads that for instance could benefit from it.

These are kind of the macro impacts that I’m hearing from you here. But when I say this to you and as an analyst, I don’t say this lightly, I admire what Aviatrix is trying to do because you’re actually… And you were very nice when you said you’re not going to cloud saying cloud is wrong. You’re just saying you can benefit from all these cloud relationships, have a multi-cloud architecture, but you can have a networking architecture as well that can drive some value on the dollar side, but also simplify the complexity. So walk us through that one time, Doug, what’s the approach that Aviatrix is recommending to get us through this chaos, enable companies to maximize the utilization of their data through networking security, but also help companies keep moving forward?

Doug Merritt:
Great summary. And, as I said, you lead and then I’ll fill in the details. It was a great upleveling. So there are a couple different components that every customer I’ve talked to is wrestling with. One of those is visibility and control. When you think about cloud, their promise was, we’re going to take care of everything for you. And we own the underlying infrastructure. They’ve invested hundreds of billions of dollars each in racks and racks of equipment all the way down to cabling and transport capability. So we manage the layer one, layer two underlay. We manage physical infrastructure, we manage everything above. So some of that networking, it just disappears. We handle it for you. That kind of work in a dev test environment. When you get to production workloads and they’re absolutely mission-critical and you’ve got a hybrid architecture or God forbid, you’ve got to have multi-cloud and hybrid architecture, control and visibility matters a lot.

Where are the packets actually going? What are they doing when they’re inside every single cloud? How do I make sure that I understand the performance? Going back to the data that could be trapped that you’re highlighting, how do I understand the performance of what my end user is feeling and how do I actually get visibility on what is happening within those clouds? So we are able to provide that because we are a data plane and a control plane and an analytics layer.

We’re a complete networking stack that runs natively within every major cloud and on those edge providers all the way back to insertion into your proprietary data centers. So we give you that uniform visibility and control, which especially as you’re saying, you’re going into a generative AI and large image model land, people have got to get much more thoughtful and intelligent about what data they’re moving where and how do I actually control that data much more effectively?

It may not make sense to move all data to one big model. You may want distribute models, you might want to keep some data close to the edge or where it’s generated. So I think that complexity with that first piece of control and visibility is only going up.

There was a blog post I put out a couple weeks ago on Embrace the Chaos. There’s a lot of the core of Splunk. Splunk has a non-structured data store. It could have chaotic data, non-normalized, non-cleansed data. Aviatrix is helping with the fact that diversity is going up.

You’re getting more distribution, more specialization. I think the number of clouds is going to go through the roof because there’s going to be a ton of purpose-built generative AI clouds to compliment the big generative AI clouds that Microsoft, Amazon, Google, and others are working to build. So we help people with that chaos because we provide a singular framework that extracts them from that chaos and gives them control again. As a CIO, how do I control my destiny? That should be very important to most CIOs. I think we’re a key lever in giving them some control over their own destiny again.

Daniel Newman:
Doug, I think you covered a lot of ground. Look, I’ve worked with you for a number of years. I watched you build a billion dollar run rate revenue company in Splunk, and I believe what you’re doing at Aviatrix is putting you on a trajectory to solve a really big problem that’s going to create another opportunity to build another billion in ARR and growth. And I’m really excited to watch that progress and watch that process, continue our conversations here. Want to thank you so much for joining on this year’s Six Five Summit. I’m pretty excited and interested to see exactly where you’ll be a year from now when we come back, but let’s talk again soon.

Congratulations on the progress and keep your head down because this is a big problem and I see a big opportunity for you and for Aviatrix.

Doug Merritt:
Well, thank you, Daniel. And hopefully, a year from now when we are back together on something like this, a lot more people know about Aviatrix and know about what we deliver as far as regaining control, enabling freedom and liberty for our CIOs out there that they can ensure control of their destiny going forward. And then we happen to drive an amazing hard dollar cost savings, which despite Splunk, we were actually 3.2 billion-

Daniel Newman:
Yeah, I didn’t give you enough credit, but you knew where I was heading with that one, right?

Doug Merritt:
Yeah, yeah, no, for sure. And I appreciate it compliment. But despite that, it was most exciting ride I’ve ever been in my life. It did not have a hard dollar cost savings component to it. And I honestly, in my entire career, have never been part of a hard dollar cost savings solution.

There’s lots of software justification you can have. We’ll help you avoid hiring a bunch of people. We’ll help you avoid a breach, we’ll help you decrease mean time to resolution. All those equate to millions or billions of dollars. Isn’t that amazing? We could do that and put 5% of your monthly CSP bill back in your pocket or 7%. Networking should be 2, 3, 4, 5% of a bill. It shouldn’t be 10, 50 and 20% of a bill. And we can help drive that equation. So I’m hoping if we can get that word out-

Daniel Newman:
I’ll end you here. It’s like in this old Xerox commercial where the guy pulled out a thing and he goes, “It’ll save me a nickel.” And then he goes, “We do X millions of transactions per day. It’ll save me.” And the point is that this is very calculatable. You can save dollars, you can get the outcome you want, improve the networking, and actually improve your bottom line. It should be a no-brainer, but the don’t know do nothing is your biggest competitors right now. So this session and so many more, Doug, let’s get the word out there and help the market see that they have optionality and what you’re doing at Aviatrix helps. I got to let you go. But thanks so much for joining us, Doug. We’ll see you again soon.

Doug Merritt:
Thanks, Daniel.

Daniel Newman:
All right, everybody back to the studio. Thanks for being here with us at the Six Five Summit.