The Networking for AI Summit | Dan DeBacker, Extreme Networks

Clips
More from The Networking for AI Summit

Dan DeBacker

SVP Product Management

Extreme Networks

play_circle_outline Role of networking in accelerating AI adoption across enterprises and data centers.

play_circle_outline The evolving landscape of networking technology driven by AI advancements, impacting various sectors.

play_circle_outline AI as a holistic enabler, impacting all areas of the network infrastructure.

play_circle_outline Importance of wired and wireless solutions for enabling edge AI use cases.

play_circle_outline Maximizing AI Workloads: The Rise of Scalable Ethernet Fabrics Over InfiniBand in Enterprise Environments

Info
Transcript

Dan DeBacker, Extreme Networks

Dan DeBacker

SVP Product Management Extreme Networks

AI is making networking exciting again. Brought to you by theCUBE Research, this exclusive Networking for AI Summit interview features Dan DeBacker, senior vice president of product management at Extreme Networks, in conversation with theCUBE’s Bob Laliberte. DeBacker explores how networks are transforming from simple plumbing to the critical foundation for AI adoption across industries. He explains why front-end access networks are central to collecting data, how Extreme’s wired and wireless solutions turn that data into actionable insights and why fabric re... Read more

explore Keep Exploring

What role is the network playing in the acceleration of AI adoption among organizations? add

What are the implications of AI on networking and infrastructure? add

What are the considerations for implementing AI within a comprehensive network framework? add

What are the key components and advantages of the network infrastructure described in relation to AI data harvesting? add

What networking technology has consistently proven to be dominant over alternatives in history? add

bolt Powered by CUBE AI

Dan DeBacker, Extreme Networks

search

Bob Laliberte

>> Hello and welcome back to the Networking for AI Summit. I'm here with Dan DeBacker, Senior Vice President of Product Management at Extreme Networks. Welcome, Dan.

Dan DeBacker

>> Bob, thank you so much. So happy to be here. Can't wait to get into this discussion about everybody's favorite topic, AI.

Bob Laliberte

>> Yeah, I think people have mentioned that once or twice, and it is certainly important because a lot of the organizations there are starting to accelerate their adoption of AI. And the cool part is the network is playing a really critical role in that. And not just as it used to be, like the plumbing, but increasingly I think it's really a true enabler of this AI innovation that's taking place and across all areas of the network. And so today, I wanted to spend some time to explore how Extreme is helping enterprises simplify operations while also building out those intelligent, adaptable infrastructures needed to enable these AI environments. So why don't we get started? As enterprises are adopting AI, in particular now we're starting to see it move across the edge, but it's also clearly in the data center and now also the WAN, how do you see networking evolving to support these new traffic patterns and use cases?

Dan DeBacker

>> Yeah, Bob, it's really exciting. I think AI is making networking sexy again. You're right, it used to be just the plumbing and it was a necessary evil, and now it is an enabler. So as we see the AI adoption curve going, and it's going so fast, the speed of what is happening and how things are changing is just somewhat mind-boggling. So taking AI across all of the entire infrastructure, so what does that really mean? Well, there's so much data. There is so much data that sits out there that has not really been harvested, and brought it back in and created information that can help a company make decisions to simplify things, to automate things, all of those types of activities that AI is now able to drive across the infrastructure. And the one good thing, you think about from an AI perspective, is that it is holistic and it's end to end. So when you look at AI, you do need to look at it from an entire network perspective. You can't just look in one little area. Oh, we can do AI ops for wireless. Okay, but that really doesn't help because I'm missing all the other parts of the puzzle. What about the WAN? What about the branch? What about the wired network? What about what happens in the data center? All of these types of things. So the pervasiveness of AI is really, the strength of it is how far out can we extend it? And so these edge networks become really, really critical in gathering all of that information.

Bob Laliberte

>> Yeah, absolutely. And now we're going to be doing some research to find out more about how AI is impacting the networks, but I'm wondering if you could comment on maybe some of the differences that you're starting to see in the traffic patterns as compared to just the traditional workloads as AI take holds, especially in these classical front-end networks at the edge and the branch where people are collecting that data but still have to maintain great experiences for the traditional networks. Sorry, the traditional applications.

Dan DeBacker

>> Yeah, absolutely. When you think about getting that information back... And so we're still early days in AI. We're just starting to scratch the surface of what's out there, what's available. So today, we look at all of that data that has to be brought back. So all that data brought back from the edge into the data center and most likely up into the cloud. We're in that first generation. So think about not only all of our applications running out to the branch, whether it's teams or it's Zoom or it's Salesforce, or whatever, you name every application out there, now that application data has to ride that same pipe and bring all of that AI data back in. So today, it's really a big centralized effort to bring that data back in, analyze that data, and then be able to make decisions on that. So we're seeing a much bigger influx of data coming from the edge into the core and out to the cloud.

Bob Laliberte

>> Yeah, excellent. And that makes a lot of sense. I think we're hearing the same thing from enterprise customers as well. I know you've emphasized the importance of the front-end access networks as really this foundation for AI at the edge. I'm wondering if you can share how Extreme's wired and wireless solutions are enabling those edge AI use cases across a multitude of different industries. I know your healthcare, manufacturing, smart cities, maybe things like that.

Dan DeBacker

>> Yeah, when we look at how we build the networks out, from a wired perspective, we really are really very bullish and have been bullish for a long time on fabric. We undoubtedly have the best fabric in the market. Couple that with a wireless solution that works very well at any scale, now those networks are really the inlets for all of that AI data. So, again, harvesting that information that sits out there in the wired network, in the fabric, in the wireless infrastructure. That's really key. And when you look at it from an AI perspective, that is the currency. Data is the currency. So being able to pull that information directly out of the infrastructure is absolutely critical. And we've done a really, really good job of being to harvest that information through APIs, and things of that nature, to bring it back in because that's really the first key. The more data that AI has, the better off AI is going to be. And so you have to feed it. You have to feed that machine continuously with a tremendous amount of data, and all of that data is coming from those front-end networks. And we're taking it a step further to actually pull data from the devices. Because pulling data from the network is one thing, that's great. I can look at the wireless infrastructure and, hey, it's working just fine. The customer experience may still not be that great. So now take it one step further and out to the edge. When we say the edge, we're saying all the way to the end device. And that end device could be a laptop in a campus. It could be your cell phone sitting in your house. All those devices that attach to the network, we want to be able to pull that information, utilize AI in such a way that we can correlate the health of the network with the health of the device, and then really provide even better information to our end users.

Bob Laliberte

>> Got it. No, I think that makes a lot of sense. And certainly the fabric that you've been developing for a number of years now, being able to stretch from data center, not just to the campus, but also to the branch, is really going to help simplify organizations as they roll out the front-end networks as well, being able to have that centralized management of that fabric. So I think that's going to help. But one of the other things I wanted to ask you about, obviously there's been a lot of Wi-Fi advancements recently between 6E and 7, and I'm wondering how you're seeing that play a role in supporting these edge environments where they're collecting data from IoT devices, other sensors and things of that nature.

Dan DeBacker

>> Yeah, 6E and really opened up the six gigahertz spectrum changed the game for wireless. It went from something that was a nice to have, kind of an adjunct type connectivity, to now most of our customers are taking a Wi-Fi first mentality to their network design. Because six gigahertz has opened up enough spectrum where there's enough data rate and bandwidth capabilities that I can utilize Wi-Fi and really never have to plug into the network from an end user device perspective. So 6E, big step forward, six gigahertz spectrum. Moving to Wi-Fi 7, even more capabilities, things like MLO and AFC, things of that nature. All of those now are unleashing more capabilities in the Wi-Fi community for our customers than to get more data back into the network. And again, we're just starting. Wait till we get to Wi-Fi 8. There'll be even more capabilities centered around Wi-Fi directly in the hardware. So there's a lot of really cool things happening over the course of the next 12 to 18 to 24 months in this space.

Bob Laliberte

>> Yeah, no, I think that makes sense, and we're certainly excited to see all those advancements and how these new technologies are going to play out and to deliver real benefits to organizations. We've brought up fabric multiple times. It's hard to have a conversation with Extreme and not talk about the Fabric. It's one of the crown jewels that you have. I'm wondering if you could talk about the Fabric a little bit more. Certainly tied in, obviously with your wired and wireless infrastructure as well, but how that is providing the segmentation, the automation and scalability that's needed to accelerate a lot of these AI initiatives at scale?

Dan DeBacker

>> Yeah, absolutely. Fabric is something that we're very proud of. It's a technology that has been around for quite some time. We started this endeavor back in 2009, 2010. So this is not a new technology, it's standards based. It is by far, again, the simplest and easiest technology out there. With that micro segmentation inherently built into the Fabric, we can really create that first base layer for our customers to make it very easy. Recently just talking to a healthcare customer of ours, they bring all these different types of devices into the network, as you can imagine, in a hospital. So you've got IV pumps and MRI machines and all these types of things, and they're constantly in testing and doing new things. Well, they can literally bring a new device in, put it on their production network, and totally isolate it from everything else. They don't have to do anything. They don't have to create any firewall rules. They don't have to do any of that stuff. It's just there. It's inherent into it. So the fabric being totally topology agnostic, plug everything in however you want to do, and having that fabric that extends across the entire enterprise network. So yeah, we started the data center. We go to the campus core, go to the campus edge. We do fabric over SD-WAN out to the branch. So have one singular fabric across your entire enterprise network, and then decide what services need to go where into what areas of that network. So from a simplicity standpoint, absolutely. Now, what happens next? Well, the ability to manage and orchestrate the fabric with Extreme Platform ONE becomes the next big step. Platform ONE built on an AI core will now be able to bring all of that information from the fabric into the AI capabilities. And so whether it's AI Expert, which is conversational, it's AI Canvas, which is a multimodal experience, or it's agentic AI, really the way you want to consume AI will be up to you and at your own pace.

Bob Laliberte

>> I think that's great, and I'm glad you brought that up because that was leading me to my next question. We've been talking a lot about the networking for AI, but we also hear a lot about AI being applied to the network itself. And so you just mentioned how you're starting to bring that into the game as well for AI ops, for automation, even doing some generative and agentic AI into your platforms. Can you talk a little bit about how that's going to simplify operations for the networking teams and IT teams?

Dan DeBacker

>> Yeah, our AI initiative, obviously I've been working very hard on it. A few things that are really key to understand. Again, pervasive across the entire infrastructure. So whether it's Wired wireless, SD-WAN, it's security with NAC or UZTNA, any of the applications, it's pervasive. So again, not siloed. When we look at AI and those capabilities, that AI for networking, we've been talking about this for quite some time, and we've had AIOps and ML, and things of that nature, have been around for quite some time. But when you think about where we're taking this and what we're doing from that AI perspective, it's not necessarily just one big thing. What we talk about here is really, it's about 1000 gains. So it's all these little things. So solve this problem, solve that problem, solve this thing. And that really is resonating with customers because they're living it today. They have issues. They have to troubleshoot a problem, they have to find an issue, they have to find things. Well, we need AI to go out there and find it for them. And so again, this is not like, oh, just turn your entire network over to AI and let it self-drive. We're not ready for that. Nobody's ready for that yet, right? That's like saying 10 years ago, get into a car that doesn't have a steering wheel and let it just drive. Nobody was ready for that.

Bob Laliberte

>> Yeah, absolutely.

Dan DeBacker

>> But if I can help a customer fix, "Hey, Bob's having a wireless issue. I don't even know where Bob is on the campus. I have 25,000 users. He's having a problem. How do I find it?" Well, how about if AI goes out there and figures it out and says, "Oh, this is Bob's problem," but I can't really fix it because AI is not quite there yet to fix the problem because it's really complicated. So instead, what AI will do is automatically gather all the information and open up a case with our GTAC support center so that the case is already opened. I don't have to ask the customer to go collect a bunch of information and we can start the process and short circuit just a bunch of back and forth before we can get into the real meat of troubleshooting.

Bob Laliberte

>> Yeah, I think that sounds great. I mean, so much of what I've heard is obviously when you're talking about the operations teams, they tend to be a little bit conservative. So any new technology, it's going to take time to adopt. I refer to this often as the time to comfort with the AI technology being able to validate it. I think that the good part is, because we're coming from a place of the consumerization of AI first, that we're starting to see, I think, organizations be more receptive to it. Not, like you said, wholesale. I'm not, "Hey, we're going to automate everything," right? It's more about how can I find those specific use cases where AI is really helping me save time? And when we've talked to organizations who use it longer, a little bit more mature in their deployments than others, it's interesting because initially the benefit is thought, "Hey, we're going to be able to find problems faster and fix them faster." And then those who are more mature and had it deployed for a while tend to be more like, "You know what? It's freeing up all this time that I used to spend firefighting to allow me to work on the more strategic initiatives." And so when I talk to end users, that's what I'm always trying to emphasize. You need to get started, you need to get comfortable with it, but ultimately understand this is going to make you more valuable to the company because it's going to allow you to focus on those strategic initiatives rather than running around all day with your hair on fire trying to put out all these problems that are popping up.

Dan DeBacker

>> Couldn't agree more. When we look at what the capabilities and what the potential capabilities of AI are about, it's not about having a staff of 20 and now saying, "Oh, I only need 10 people." It's having a staff of 20 that really has the productivity of a staff of 50. That's what we're going after. And I think that's the mindset that we really want our customers to adopt is that we're not out to eliminate jobs. And there's so much in the press about all this stuff and the FUD around AI and "Oh my goodness, what's going to happen?" But to your point, it's exactly that. It's how do we create more efficiency? So again, take that staff of 20 and make them work like a staff of 50. How awesome would that be?

Bob Laliberte

>> Yeah, absolutely. And it's just like I said, it's not going to happen overnight, but the sooner you get started, the sooner you're going to get to achieve that level of efficiency and be able to drive more real results for your business. So I think that's going to be great. Last question, I have to ask you about this. This is one of those classic InfiniBand versus Ethernet. Why does Extreme believe ethernet fabrics, combined with your architectural approach, will provide a more open, scalable and enterprise-friendly path?

Dan DeBacker

>> Well, I think you just hit it right there, right? Ethernet has won the war over and over and over. You go back many, many years and there was FTDI and there was Token Ring, and there was DECnet, and you name it. All these things. And ethernet won the war. It's scalable, it's simple. It's well understood and well known. And when you think about how we're going to do networking moving forward, and you're creating networking for AI inside your data center, you already have all of your people already trained. They all know ethernet. And networking for AI, there's really nothing special. You can run it across an IP fabric, BGP, EVPN, VXLAN tunnels. You do some storage networking capabilities, RoCEv2, things of that nature, it's all things that we're doing today. So why would I try to bring in some different technology where it's more expensive and I'm going to have to go train all of my people around InfiniBand, and it's just for one specific area inside the data center, as opposed to saying, "Well, I use ethernet everywhere, so I understand ethernet. My team understands ethernet." And really the instantiation of a data center network for AI and AI workloads is ethernet and it works just fine. So I think from that perspective, it's going to come down to that return on investment and how hard will it be for me, as a network administrator, to go find people who understand InfiniBand? And then I train them up on how hard will it be for me to keep those people? I mean, adding more different types of protocols is really counterintuitive to what we've been doing for the last 20 years.

Bob Laliberte

>> Yeah, no, I think you're absolutely spot on. I think someone had told me they went to a talk given by Bob Metcalfe, the founder of ethernet, right? The father of ethernet. And he said, "I don't know what's going to come after ethernet, but I'm pretty sure it's going to be called ethernet."

Dan DeBacker

>> Yeah.

Bob Laliberte

>> And I think he's been, so far, he's been right. Hey, Dan, listen, I want to thank you so much for sharing all your insights on how Extreme is advancing networking for the AI era. It's really clear the network, it's not just becoming an enabler of AI, it's also becoming an accelerator of business value as well. So Extreme's vision of a universal cloud native AI powered network, I think that really highlights how customers can scale AI across various industries without adding complexity. So as AI is reshaping every business, the need for intelligent, secure, and adaptable networks has never been greater. Thank you so much for joining me today.

Dan DeBacker

>> All right, Bob. Thank you very much.

Bob Laliberte

>> Yeah, great to have you on board. And everyone, stay tuned for more Networking for AI Summit to come.