 Live from Barcelona, Spain, it's theCUBE. Covering Cisco Live Europe. Brought to you by Cisco and its ecosystem partners. Hello everyone, welcome back to theCUBE's live coverage here in Barcelona, Spain for Cisco Live Europe 2019. I'm John Furrier, my cohost, Stu Miniman. Dave Vellante's out there as well, cohosting this week. Our next guest is John Apostolopoulos, who's the VP and CTO of the Enterprise Networking Business Unit, lab director for the innovation labs. He had to talk with us about AI, some of the great innovations. John, thanks for coming on theCUBE, great to see you. Thank you for inviting me. Pleasure to be here. So Cisco, obviously big big announcements, the messaging's coming together, certainly the bridge for the future, bridge for tomorrow, whatever the phrase is. You know, kind of looking at that new world, connecting on-premise, cloud, ACI anywhere, HyperFlex anywhere, a lot of complexity being abstracted away with software, separate from the, a couple from the hardware, a lot of scale in the cloud, and IoT and all around the edge. So software is a big part of this. So you can't help but think, okay, complexity, scale, you see Facebook using machine learning, machine learning and AI operations now, a real conversation for Cisco. Talk about what that is, how are you guys looking at AI and machine learning in particular? It's been around for a while. What's your thoughts on Cisco's position and opportunity? Sure, yeah, Cisco's been investing and using AI for many, many years. What happens Cisco, like most companies, we haven't really talked about the machine learning as a term because machine learning is a tool we use to solve different problems. So we've talked about what are the customer problems we have and then how we solve them and how good our solution is. But we haven't really talked about the details about the how, but we've been using at Cisco and like myself in past career and so forth for many, many years of machine learning. Security has been using it for multiple decades, for example. And where's the use case for machine learning? Because it's one of those things where there's different versions and flavors of machine learning. Machine learning, we know powers, AI and data fees, machine learning. So do you have all these dependencies and all these things going on? How do you, how should someone think about sorting through machine learning? Well, machine learning itself, that term is very broad term. It's almost as big as computer science, right? So that's where a lot of the confusion comes in. But what happens is you can look at what types of problems you're going to solve and when you try to look at what types of problems you're going to solve, some of them, for example, some problems you can exploit the fact that there are laws of physics that apply. If the laws of physics apply, you should use those laws. We can either figure out that if we drop this, this will follow some speed by measuring it and using machine learning, or we have gravitational force and friction with the air and we account for that and figure it out. So there are many ways to solve these problems and we want to choose the best method for solving each one of them. And when people think about Cisco, the first reaction isn't like, oh, machine learning, innovator. What are you guys using machine learning for? Where has it been successful? What are you investing in? Where's the innovation? Sure, sure. So there's a lot of problems here that come into play. If you look at the customer problems, one example is all the digital disruption. We have on the order of a million devices, new devices coming on the network every hour throughout the world. Now, what are those devices? How should you treat them? With machine learning, we're able to identify what the devices are and then figure out what the network posture should be. For instance, for an IoT device, we want to protect it, protect it from others. Another big topic is operations. As you know, people spend, I think it was Gartner identified that people spend about $60 billion per year on operations costs. Why is it so much? Because mostly operations are manual, about 95% manual, which also means that these changes are slow and error prone. What we do there is we basically use machine learning to do intelligent automation and we get a whole bunch of insights about what's happening and use that to drive intelligent automation. You may have heard about Assurance, which was announced at Cisco Live one year ago at Barcelona. And both in the campus with DNA Center, we announced Cisco DNA Center Assurance and the data center we announced, network analytic engine. And what both of these do is they look at what's happening in the network, they apply machine learning to identify patterns and from those patterns identify is there a problem, where is the problem and what's root cause and then how can we solve that problem quickly. John, can you help us connect where this fits in a multi-cloud environment? Because what we've seen in the last couple of years is when we talk about managing the network, a lot of what I might be in charge of managing is really outside of my purview. And therefore, I could imagine something like ML is going to be critically important because I'm not going to be touching it, but therefore I still need to have data about it and a lot of that needs to happen. Yeah, well one of the places ML helps with multi-cloud is the fact that you need to figure out which, where to send your packets. And this comes with SC1. So SC1, we often have multiple paths available to us. And let's say with a move for Office 365, people are using the SaaS service and they want to have very good interactivity. One of the things we realize is that by carefully selecting which path we can use at the branch and the campus too, we can get a 40% reduction in the latency. So that's the way we choose which COLO or which region or which site of Office 365 to send the packets to dramatically reduce latency. What's the role of data? Because when you think about it, moving a packet from point A to point B, that's networking, storage acts differently. You can store data. Data's got to come back out and be discovered. Now if you have this kind of horizontal scalability for cloud edge core coming into the middle, get up the data. So machine learning needs the data, good data, not dirty data, any clean data. How do you see that evolving? How should customers maybe think about preparing for either low hanging use cases? Just what's your thoughts and reactions on that? Yeah, well the example you gave is a very interesting example. You described how you need to get data from one point to another. For instance, from my device to a data center where the applications are over the cloud. And you also mentioned how the many things in between. What we care about not necessarily, it's not necessarily the application data. We care about, you know, we want to have the best network performance so your application's working as well as possible. In that case, we want to have an understanding of what's happening across the path. So we want to pull telemetry in all kinds of contexts to be able to understand, is there a problem? Where is the problem? What is it and how to solve it? And that's what assurance does. We pull this data from the access points and the switches and the routers. We pull in all kinds of contextual information to get a rich understanding of the situation to identify if there's a problem or not and then how to solve it. It's the classic behavioral contextual paradigm of data. But now you guys are looking at it from a network perspective. Exactly. And as the patterns change, application-centric programmability of the network, the traffic patterns are changing. Hence the announcements here about intent-based networking and hyperflexed anywhere. This is now a new dynamic. Talk about the impact of that from a AI perspective. How are you guys getting out front on that as it's not just North, South, East, West. It's pretty much everywhere. The patterns are could be application-specific at any given point on a certain segment of a network. I mean, it's complex. Yeah, it's complex. One of the really nice things about intent-based networking though is this fits in really nicely. And then that was by design. Because what happens is intent-based networking, as you know, a user expresses some intent or something they want to do. I want to securely onboard this IoT device. And then it gets activated in the network and then we use assurance to see if it's doing the right thing. But what happens is that assurance part, that's basically gathering visibility and insights in terms of what's happening. That's using machine learning to understand what's happening in the network across all these different parts that you mentioned. And then what happens is we take those insights and then we make intelligent actions and that's part of the activation. So this, with intent-based networking, this feedback that we have directly ties with using the data for getting insights and then for activation, for intelligent actions. John, I always want to get the update on the innovation lab. Is there anything particular here at the show or what's new that you can share? Oh yeah, so we are looking at extending IBM to the cloud, to multicloud, to multiple devices. So there's a lot of really fascinating work happening there. I believe you're going to be talking to one of my colleagues later too, TK. And he's, I think, hopefully going to talk about some of the machine learning that's been done. And that's already productized, as you know, in encrypted thread analytics. And that's an example of where we use machine learning to identify if there's malware in encrypted traffic. Which is really a fascinating problem. That is hard. I mean, I'm looking forward to that conversation. So some members of Cisco, Dave McGrew, in particular, Cisco fellow, started working on that problem four and a half years ago. And because of his work with other colleagues, he was able, and they were able, to come up with a solution. So it was a very complicated problem, as you saw. But through the use of machine learning and many years of investment. Plus the fact that Cisco had access to TALUS, which has, you know, they know the threats throughout the world. So their list of data in terms of all kinds of threats is massive. The volume of that's where machine learning shines. I mean, you're seeing the amount of volume of data coming in. That's where it could do some heavy lifting. Exactly. And that's the point of Cisco's strengths. The fact that we have this massive view on all the threats throughout the world that we can bring it to bear. Yeah, network security foundationally just creates so much value for apps. Final question for you. For the folks watching, what's in your opinion the most important story here at Cisco Live Barcelona that people should be paying attention to? Ah, so I think how we're trying to extend across all these different domains and make it like one network for our customers. This is still a journey. It's going to take time. But with intent-based networking, we can do that. And we're going across campus when data center to multicloud. And how hard is cross domain? Just put it in perspective. Cross domain, traversal, and then having visibility into these latency from a physics standpoint. How hard is it? It is, it's quite hard. There's all kinds of technical challenges. There are even other sorts of challenges. This is Wi-Fi, right? IEEE 802.11 defines the QOS standard for wireless. And that's completely different than how the internet group, IETF, defined it for wired. So even between wireless and wired, there's a lot of work that has to be done. And Cisco's leading that effort. And having all that data. Great to have you on, John. Thanks for spending the time and demystifying machine learning and looking forward to this encrypted understanding with machine learning. And that's a hard problem. Looking forward to digging to that. Again, truly, the breakthroughs are happening with machine learning. And adding values. This is what application-centric world. This, the data is all about the data. This is theCUBE. Bringing you to the data from Barcelona. I'm Jefferson Stu Miniman. Stay with us for more coverage after this short break.