 Live from the Moscone Center in San Francisco, California, it's The Cube at AWS Summit 2015. Okay, everyone, we are live in San Francisco. This is The Cube. This is our flagship program where we go out to the events and expect the sizzling noise. I'm John Furrier, co-founder, still getting on meeting my co-host. Mark Farley, book author, expert guru in all things tech. Our next guest is Guru Shahal, who is the guru at VP of Products at Abbey Networks. Welcome to The Cube. Thank you very much for having me. So I got to ask you, Abbey Networks, it's a networking play, but app delivery we're going to talk about, but Amazon here is just unloaded and more truckload of goodness in their typical fashion, more stuff, machine learning. We got some more stuff across the board on the mobile's piece, containers. So it's a developer-focused kind of environment, very DevOps, very cloud, we love that. But at the end of the day, the networks are the bottleneck. So give us an update on what you guys are doing as a company, and then what's the action in the network layer? And what does that mean for developers and for customers? Right, John, thanks for having us. We're pretty excited to be here at AWS. Abbey Networks got its start a few years ago, and we've really built a next-generation application delivery platform that essentially combines industry's first scale-out load balancer with integrated real-time monitoring of applications and end users. And these two things are important because the combination of these two results in a very dynamic application delivery environment that can track application end user experience and continually deliver the application with the highest possible performance through end users. And when we got our start, we really spoke to several hundred customers over the course of about a couple of years, and we asked them, how are they handling the three big trends? The adoption of cloud, the consumption of applications moving to mobile, and then finally the changing application architectures. You were talking to our developers here. You walk over to any developer, and no developer is building monolithic applications anymore. They're building microservices. They're building highly fragmented, small services which then work together to bring together an entire application. And the question we asked ourselves is, what is the impact of those three big changes, cloud, mobile, microservices on the network? That was the genesis of the company, and the impact was twofold. As the workloads move to cloud, as the consumption of the workloads moves to mobile, as the workloads themselves become highly fragmented and distributed, the importance of the network is twofold. The connectivity has to be paramount and high performance because naturally you're more and more separated from where the workloads are. But more important than the connectivity, which is essentially the plumbing, the network itself has to be more intelligent. It has to be aware that the application is not just this one big monolithic application sitting inside a VM or a big honking box, but it's a set of 50 microservices working together. It needs to be aware of the fact that the consumption has moved rapidly from that laptop here to an iPhone, to an Android phone, to maybe somebody in Russia, from somebody in France to somebody here in the Bay Area access in the application. So the network has to be aware of the consumption, the location of the application, and then continually optimize and connect the two in a way that end user experiences continually optimize. Now this goal has been in the market for a very long time and you'll find there are specialists and specialist technologies that enable you to meet this goal today. And what Avi has really done is we've taken a lot of the complexity out of the system and build an amazingly intelligent platform that combines that software-based scale-out distributed load balancing aspect with real-time monitoring of end user experience and applications. And then in response to that can just continually optimize application delivery. So that's... You have an integrated analytics engine for that, for processing all that information in real time. Mark, the technology is something I salivate. I mean, technologists, you know, I salivate over it all day long. Our customers love it. This is truly, you know, people have been talking about big data for a very long time. People have now begun to talk about big analytics. This is one of the first true applications of big analytics to a real-world problem. How do you take a human out of this equation, a specialist human, of looking at hundreds of telemetry data points and then adjusting and tweaking your network to adjust, to deliver the application in the best possible fashion? That is key, that once we do that, we take so much, we democratize this technology in a way that so many more developers, so many more enterprises can take advantage of that. So what we've done is we've applied STN principles. You've got a distributed data plane, a distributed load balancer. These continually collect over 500 data points every second of telemetry for every application, every microservice that they serve. So you've got 50 microservices sitting around, they're collecting 500 data points, but five to 700 data points continually, sending it to scale out analytics engine that then processes all of this data in real time and takes those analytics, provide those for consumption for the enterprise, for the DevOps consumer, for the operations guys, but also takes that and feeds that right back into the data plane. So it's going to go and say, boy, I'm running out of capacity, let me scale myself out. Or maybe the application is running out of capacity, let me go and scale the application. And all of this is happening in real time. So I want to get your take on it. I'm sorry, did I interrupt you? No, no, go ahead. A comment I made earlier in the keynote live tweeting, I wrote, it's pretty clear that enterprise IT is a long secular decline, that the cloud is going to win as the primary modality of delivering enterprise computing. I know this offends a lot of people, but it would be a good time to come to terms with this reality. And then the threads come in. The question is, how does the enterprise get to the cloud? And I said three possible answers, the internet. This terrifies the enterprise for variety reasons. You know, two carriers cloud interconnect services, conventional carriers or other startups looking to launch like IIX and whatnot. And then the single pipe to the enterprise notion, hearing all this is crazy. And then co-location cloud interconnects are the classic equinics of the world. So give us the take on this, because this is important, because this now gets down to old school, new school. What side of the street will the future be on? So how do I get to the cloud? Do I use carriers? Do I use an interconnect startup or priest and co-low? These are huge issues, right? I mean. Sean, such a good question. If there's one thing that people have learned over the last 10 or 20 years as an IT, it's not to bet on one thing. Because the moment you commit to one thing, that your problems start there, it might be the best choice today, John, but I've been attending AWS summits for the last probably three years. Every summit I'm surprised by the pace at which public cloud computing and cloud computing in general is evolving. I can't tell you what's the right answer today. Nobody can, because all three of these are evolving at the best pace. The right thing for enterprise IT to do is to ensure that the building systems that provide them single points of control and visibility across all three methods. So they can in real time go, well, this is not working out. I'm not getting a great deal today. Let me move from a direct interconnect using Equinix to perhaps the standard internet with some sort of an overly security solution and connectivity solution into what Cisco might provide with InterCloud Fabric, for instance. Or maybe that's not working out for us. Now I want to go and just co-locate right next to Amazon. Or maybe I have a dedicated pipe that Amazon wonder provides to me. So there's all these different options, but what you'll find is as soon as enterprise locks itself into one, I guarantee it within 12 months, it's not the optimal option for you. So the right thing to do- So there's two things you're saying. One is don't lock into one thing, because you get stuck there, you get stuck in the mud, all kinds of bad things can happen. The wheels come off the bus, whatnot. And two, the things we just talked about are evolutionary, they're evolving. That's right. So I got to ask you, as someone who's in the trenches and looked at a lot of the ecosystem in Silicon Valley, where's the action? I mean, so we're seeing Google, for instance, doing a lot of peering, my guess is, they want to kind of get their own interconnect going where they can have private connections or circuits or whatever you want to call it into the enterprises as an on-ramp, if you will, to the cloud, right? Overlaid. And avoid the service providers altogether. So where's the wind blowing in this area? Is the trend to have not a lot of network hops? Is the interconnect an issue or not? It absolutely is. And you know what I say? I tell people, follow the money. And if you follow the money, the amount of fiber that's been laid out over the last few years, the amount of new interconnects that are going down, the amount of latency that people are focusing on taking out of the system, that's where the money is being spent. And again, it's very intuitive. You've got to go back to core principles. If you're, if where you deploy application changes to being highly distributed out in the cloud, if how you consume it changes to very dynamic, jumping from mobile, from one continent to another and so on, and how you design your application, the application architecture changes, what's the most important thing that impacts your application performance today? The network. And that's where the focus is. And eventually the winners are going to be based not just on the level of service they provide and the services they provide, like AWS provides, but on the end user experience. I can relate a very quick story for you. 99% of the conversations around cloud are TCO conversations. Hey, help me move to public cloud, Amazon, and tell me how much money I'm going to save and how much quicker I'm going to get. And yet there's an entire section of enterprise that finds that completely uninteresting. Because if you live or die by end user experience, if you are coupons.com, if you're a Mobi save, if you're a Mercado 3, if you're any one of these, any one of these digital experience companies. It's more costly to break the other side. There you go. It is much more expensive to lose even a second of performance, a millisecond of performance than it is what you gain from going to public cloud. And that's where the final is. So just to find my previous sunk cost to making sure I don't lose money on the top line. That's exactly right. Top line is becoming the new game. It's not just about bottom line, but people are saying, A, prove to me you're not going to be costing me top line. End user experience is not going to go down. So that requires network investment. And the other one is, of course, how can you have me increase top line? All right, so the next level deep I want to go with you on that is, okay, now software models are changing. Software defined, software defined networking. So if we have a world where the network is programmable north, south, east, west. And this is evolving whatever shoe drops on that. You're going to have policy based something, right? So what is happening in the software world? How much virtualization is impacting that? What are the new software models you're seeing? What are some of the innovations that are blooming out of the startup world or in the big comes like Cisco, whatnot? What's happening in that area? That's going to be the real key. It's a great question, John, because what's really happening is people are borrowing innovation from one part of the market and now applying it to other parts of the market. It's amazing. This trend is so universal. So if you look at applications, back when I was in school, you were talking about monolithic applications. They rapidly evolved to multi-tier. Still two-tier, three-tier, maybe four-tier applications, right? And now the wage is services, microservices. The service-oriented architecture has been around for a while. I mean, all Amazon services are building services. So this idea of small, componentized application services that can scale out on demand. That idea is there in the application space and the new true innovation is the application of that into infrastructure, into the network. How do you take this idea of rapidly scaling microservice-oriented architectures where I can press the button and my policies are centralized, but the actual impact of that, network services, the network functions are established on demand as microservices close to the application microservices themselves. That trend is now seeping for our applications into the infrastructure. And what people really talk about is how do you mirror what is going on in the application world into the infrastructure world? So you guys have a lot of smart guys. A lot. You've got a lot of smart guys. Way smarter than me. And so one of the interesting challenges you have is trying to stay aware of what's going on in all of this development and application space, open source technology. My guess is you have a lot of open source technologies embedded in your product. A lot. Right? And so the question I've got for you is do you spend more time combing and trying to figure out how to integrate these other things that are already done or do you invent new microservices or some of both? It has to be a sum of both, always. So the goal for any startup, for any company that is delivering innovation is twofold. One, innovation is not just glue logic. You can't take four open source technologies and combine those together. Your customer can do that too. Customers are also getting equally smart. So if you could do it with 10 engineers, I'm sure a customer will say, well, maybe I can hire 10 contractors and get this done as well. True core innovation is core IP that you have to produce, that our engineers have produced. So for instance, if you look at Avi, I'd say about 60% of the resources are focused on true core innovation. How do you take an incredibly hard problem, state distributed at scale across millions of components? How do you take millions of data points, billions of data points and process those in real time to provide analytics, machine learning, all of those? How do you feed that back into the data plane? So there's about 60% of our workforce that focuses on that. And then we've got amazing experts in open source technologies that say, okay, we don't need to build this. This is context. It's not core. This is context that's pulled from here. If you need to make changes, let's make the changes and contribute it back to the community. So we're very focused on that. So we've got, much like any other startup, a GitHub repository, you can go there. Every change we make to an open source component, every improvement we make, it gets contributed right back to open source. So it's a give and take, it's a give and take. I got to ask the question, because I know you work at Cisco. Is that true, you work at Cisco? Yes, we got, much like almost anybody else in networking, I've done my tool of duty. And you're posse over there at the AVI network. So all Cisco alum. So I got to ask the question because in everyone I know that works at Cisco, they're super smart in networking. But they always, the phrase that's been kicked around for over a decade is, we got to move up the stack. And it's been kind of an internal discussion at Cisco because you've owned the networking. I mean, many people have tried to compete with Cisco. Just got decimated, the nested nest, so you just entrenched like a tick in all businesses. You know it's hard to rip Cisco out. But the pressure's always been to move up the stack. So the question is, has Cisco moved to the stack? And what are people moved up the stack already? We're hearing this is up the stack. Amazon is the stack. So the apps are focused. So the old notion was, I got my portion of the stack, my networking, I'll leave everything up and move the policy base, whatever. What is the shift there? And what do the old incumbents have to do to either move up the stack or equivalent? I think philosophically, John, if you take a step back, Cisco, whether it's VMware, whether it's HP, whether it's any other large established legacy vendor, they have to move up the stack. It's just the definition of move up the stack is very different based on who you're talking to. Everybody will obviously define it to their own interests. Now, Cisco, I think over the last five or six years has made some very interesting moves. So I ended up at Cisco through the acquisition of Nuova Systems, which then gave Cisco an insight into the server side, into the compute, which was a big missing piece with UCS. That thing is, UCS is a three plus billion dollar business today, if you look at the last quarterly call. It's amazing business, not only is it important that you have a leg in that piece, but it is important that you begin to understand how customers are deploying applications and what are they doing with these applications. So now I can see Cisco pulling together their compute offerings, their networking offerings, startups like MetaCloud that they bought a few months ago, which then give them this full stack, full service offering where they can walk into a data center and say, you know what? From an overall data center cycles, compute cycles, networking cycles, all the way up to how do you deploy an application? I can provide this whole thing for you. That space, which popularly is now called hyperconvergence, I think every single vendor is going to try and own that space. It is the right place to own, not just because customers no longer want to be their own SIs. They don't want to be a system integrated. They want somebody to just rack and roll, press a button, and I've got, you know, something that- I've got to ask you a question. If I'm your best friend, we're having a beer, we're talking. I say, Guru, look, I don't have a lot of time going on in this networking space. I just want to recognize failure and winners, winners and losers. What does the bad decision look like in terms of like, let me say company name, but like environment, the kind of decisions I make, the kind of product, the kind of business models in your area that you're playing in. What's the, you know, what does it look like? I want to recognize and stay away from the bad guys and go with a modern, scalable, horizontally scalable, open source, whatever it is. What does it look like to be a winner and what's it like to be losing? You know, there are two trends we see very clearly on the bad decisions part. And because I can't predict the future, I'm sure one day I might live to regret these examples, but there are two examples. You can't think it back, it goes live. Yeah, there you go. So there's two examples I can take from my recent past. One is when people start investing in technology for technology's sake. And you know, this is not probably the best example, but if you take some of the work that came out of academia in networking, technologies like OpenFlow and others, continue to develop those without focusing on the end user problem and use case you're solving, can be an issue. So I think some of the customers who got really carried away with the hyper round SDN way back two, three, four years back have spent a lot of money and time and effort into those. Now, of course, it's a different word. You've got much better products that are addressed. But that was a good bet though. SDN was a good, maybe a little early premature. SDN was a good bet. But again, how you do it and the vendors and products that you bet on is very, very good. Bolting on. John, I think a more interesting area and relevant to Avi might be talking about application delivery. Yeah. Right, so what happened in the application delivery space? How could you see what was going on with winners and losers there? Right. Because it's no to you. That's my second example, which is that if you look at sort of this SDN whitewashing or cloud like washing around what's going on with vendors, I can tell you every single vendor in the application delivery space over the last 10 years has done only one thing. They have made it a speeds and feeds game and they've moved the software that they used to put in sheet metals and put it inside the VM and say, ta-da, we're ready too. Here's a VM. We adopt cloud, it's awesome. Oh, you want to run us on Amazon? Not a problem, I've got a VM for that and an AMI image. Boy, I got to tell you, I talked to probably at least 50% of the Fortune 200 I've spoken to personally, right? And every single person says the same thing. Just taking something that used to reside in a server and moving into a VM is a terrible approach because all that you do is your operation complexity goes way, way high. What you really have to do is apply SDN principles, central policy depositories, central point of control and visibility and distributed microservices that can then be applied no matter where the application goes. So we got coming up on time here, I want to ask one final question. If you have a final question, we can get to that too as well. Web services finally is here. I mean, I look at that chart that Andy Jassy put up there, I'm like, hey, that's service-oriented architectures, web services, that's all the stuff that was kicked around in 2000 timeframe. S-O-A. S-O-A, oh my God, it's Lego blah, oh yeah. So it's here, I mean, everything kind of has their time and gets adopted. What does that mean for us, this world that we're in? Is Amazon completely drinking their own Kool-Aid too hard? Will they ever win the enterprise? Will they get a piece of it? I call it the big way that's going to hit the beach. They'll get it some erosion, but how deep they go is a function of how fast they can get that amplitude of that wave going. So the question is, how far will Amazon go with their model and who will stop it? Web services is absolutely real. I am talking in mid-market, e-commerce, SaaS players today who are adopting microservice architectures and web services over all, general and service-oriented architectures very, very aggressively. The real question is not how successful Amazon or how far Amazon will push out. They will continue to push out. They have a great capitalistic reason to do that. I think they have leverage. I think the more important thing to understand is that the ideas are out there now. So whether it's Amazon, whether it's somebody else, or whether it's a private cloud infrastructure within the enterprise, the ideas are out there. Microsense is at real. Application architectures are changing. That change will drive infrastructure change all the way down the networking stack and how we do business and how we connect applications to customers. That's what we are betting on. We are seeing that in the marketplace. Our customers are telling us that on a daily basis and it's an exciting time. So tell us quickly how many employees you guys have. What's the funding? Who's backing you guys? How much revenue you have? So we launched in December last year. Our backers are, we've raised two rounds of funding from Greylock, LightSuite Venture Partners, and Menlo Ventures most recently in our series B. We are all based down here in Sunnyvale except for our sales and marketing team which are much more broadly placed. We are several dozen people now, over 50 people at this point. We've had customers since about the middle of last year. We've got about over two dozen customers today in various phases in production, in lab evals and so on and so forth, paying customers, and these are largely focusing on Fortune 200 at this point. So very, very focused on these. Who's the partner at Lightspeed? Barry Eggers, who you know very well. And you guys, they did a great deal with Nutanix as well. They did Nutanix, AppDynamics. So you see that common thread, you just touch upon web services. All of these companies are betting on a new application architecture, whether it's Nutanix with hyper-convergence, AppDynamics with highly distributed architectures, Avi Networks with taking that same microservices philosophy and applying those to the network. We're all betting on the same trend and we're- We're all integrating them too, making them all engineered. Yes. All right, we are here with the guru. He's actually a guru. That's his name, guru Chahel, VP of product at Avi Networks. Hotstar, keep an eye on these guys. We'll certainly, we will be keeping an eye on them. Great conversation. Thanks for joining us on theCUBE. We'll be right back after this terrific live here in San Francisco at Amazon Web Services Summit. I'm John Furrier with theCUBE. We'll be right back.