 Hi everybody, this is Dave Vellante, and welcome to CUBE Conversations. We're going to talk about modernizing network infrastructure with a friend of ours, Rich Napolitano, who's the CEO of Plexi, I'm joined by Paul Gillan. Rich, it's great to see you again. Great to see you, David. So it was about a year ago, a little over, we were down at Big Data NYC, you came on, you had just taken over as a CEO of Plexi, you popped out of the EMC after a great career there, obviously son, long time, New England technology executive. So give us the update, how's Plexi doing? What's the first year been like? No, well thank you, it's great to be here, and spend some time with Paul and you, as always, it's always a fun conversation, so I'm looking forward to it. It's been a great year. We, we, end of last year, we released our product as the basis of our momentum this year, we've talked in the past about growing our revenues 10X this year, and we're on track to do that, so I'm really excited about that. Team's been working really, really hard. As a startup, I think, you know, you gotta simplify things that are complicated, so, you know, I focus on three things every single day. First is, you raise capital, and we'll talk about that. You gotta deliver products, and you gotta sell those products, and you know, my day consists of working on those three things. So, we shipped, at the end of last year, our 2.2, 2.1 software, and our new Switch 2 platforms, and that became the basis of our revenue expansion this year, and we're making great progress there as some more key customers, and great momentum on that front. In addition, we announced in the summer that we signed Arrow as our distributor, and Arrow working very closely with them to recruit a new kind of partner. We call them cloud builders, which we'll talk more about, I think, later in the conversation. And these are partners that are helping customers build public and private cloud infrastructure for the future. And so, you know, deliver products, build out your go-to-market, and raise capital. And in August, we announced a $35 million round with some new investors, and to really enable our expansion around our go-to-market. So, talk a little bit more about Plexi. I mean, everybody talks about traffic, network traffic transitioning from North-South to East-West, but you guys take it a step further. I mean, I kinda look at you, I think. DevOps for networking, but talk a little bit more about how you're different there. Yeah, I mean, DevOps networking is not a bad way to think about it. The requirements in this new era, and we talked about this, I think the last time I saw you were shifting, and we see that very clearly now, and the market's clearly coming to us. So, part of it is the nature of traffic. It's really the nature of applications have changed. And for 20 years, applications were client-server oriented and you'd send a request down to a client where you get your result and that resulted in North-South traffic. More and more applications scale horizontally now or scale out. And so, you see more traffic going East-West, whether that's big data environments where Hadoop and things short out requests or whether it be VMware environments where you have more and more VMs connected and then you have operations that are causing cross-network traffic like VMotion. Things that cause the network to work both North-South and East-West when the prior architectures were really designed for North-South traffic. But it's not just about traffic. The big learning this year for us is that these cloud builders, they want to build public and private clouds that are tremendously agile. And the traditional network architectures are very, very rigid. Traditional networks are really, when they're implemented, your architectures locked in. We kind of jokingly call them wire-defined networks. They are defined by their wires and their cables. The day that you cable it, your architecture is determined, your connectivity is determined, your bandwidth is determined, and your bandwidth is locked in that cable. And in our architecture, we would say that our architecture is just more flexible. Our architecture is, we generally don't use SDN because that has too many load means. We would say that our network is a network that can be defined by software. In some respects, the opposite of SDN, it's NDS. So our network can be defined by software. So if you're gonna have a network that can be defined by software, then you have to have a founder, Dave Hussack, awesome guy, MIT guy, figure this out. And he loves saying, if you're gonna have a network that's defined by software, you'd better have a definable network. In other words, a network that's not defined by its cables. So what you're talking about, it's gotta be a little scary for a enterprise IT guy who's got a big North-South architecture, as you're talking about. I mean, those apps are not gonna change. The structure is not gonna change. How are you fitting into this structure? So the key thing about our product is that it's both future-proofing and kind of new architecture and compatible with the legacy. And I think the world, having been in technology now for over 35 years, we declare the end of eras very, very quickly. And the tale is enormous as we know. And so any new technology needs to embrace the encompassing by some means, especially any other parts. So our product has those properties, which is southbound, we are a very traditional layer three, two, one networking stack. So we are fully compatible. I mean, we use Broadcom chips, which really define the standard for interfacing. Whether it be Cisco or Rista, you name the company, they all use exactly the same components. And so the very traditional networking protocols we support. So southbound, you can plug our networks into existing networks and cap your investment in that and then grow in a much more agile way. Because it's not just about east-west, it's really about agility. It's really about allowing people to build, say, private clouds that are as easy to use and consume as public clouds. But you've got apps that are built on this north-south architecture and you want to get them to the new world. I mean, is that a, it sounds like an unnatural act. And it's your app driven. What's your role in making that happen? So even these traditional north-south oriented applications as you virtualize them, you cause problems in these environments because many of these apps were written in eras when the infrastructure was totally dedicated to them. So now you put them in a virtualized environment and you have many different applications on the same infrastructure, except you're using the same static network architecture. So as you virtualize these applications, what you can cause is hotspots in your network. So what our network would do is looks at these applications by integrating into them. So part of the ah-ha, Plexi, is become application aware. If you read a lot of our material, applications-centric networks is often we talk about. And so we actually look into these APIs of these operating environments like, say VMware or salesars.com or Hadoop or other applications and understand the nature of these requirements. And then what we call fit and render. We fit, we calculate the optimized topology for these applications. And then we implement that topology dynamically in a network. So there's a lot there. So you're talking about asset leverage on the existing infrastructure. And you're talking about being able to modernize the application portfolio without ripping and replacing it. And then you also mentioned application-centric. It's a term that Cisco uses. So juxtapose what you guys are doing versus the big whale in the business. It's actually very interesting. Even just recently, a lot of the analysts are picking up that what we have built is a very similar ACI. Except the difference is that the ACI is really built on what we would call a platform 2-oriented network architecture, traditionally conspiring, or a wire-defined network. So they haven't fundamentally changed the underlying substrate of the network. They're solving a very complex application-oriented problem. We would say that you need both. In fact, we're actually very friendly to things like NICERA or other overlays. We don't want to be the overlay. We want to be a great underlay to enable, whether it be Alcatel's virtualization or VMware's network virtualization, we're very friendly to those environments. Because it turns out that our network is so dynamic that these overlays can tell us what topology they want. And we can change that topology. To be specific, in a VMware environment, we can understand that certain nodes in the VMware are actually part of one cluster. And we can create a topology that says, these are all grouped together. And so how do we do that is we have integrations into VMware or potentially Nuage, where we look at the requirements for these grouping, same in Hadoop. And then since our network is defined not by cables, but by photonics, so we can shape our physical layer by using, it's not really mirrors, but conceptually mirrors, that basically say that this group of nodes is on red, you're on blue, you're on green, I'm on purple. And on one physical network, we can actually be talking to each other separately and discreetly. And I can change that once a second if I want. It's as if I group and fit and render, you got it. So that's what you're calling a platform three technology, but you're not, you're able to plug in to the platform two infrastructure for you, like a 2.5 or? So we're ready? Where I, it's actually, I think Tuchio often uses 2.5, which we're kind of fond of as well. But so we can bridge, you need to bridge, right? Because you need to bring people forward. Everything just can't be ripped and replaced. So the fact that we integrate into VMware, where if you really think about it, if you go back to the early days of VMware, when EMC acquired them and I had just joined right shortly after that, VMware was very successful in test and dev. And the moment that Intel created multi-core and the instruction said, then VMware really accelerated. Because effectively the underlying substrate of compute became very favorable to the VMware environment. We would say that we're exactly the same thing to the network because the underlying substrate of the network now in traditional servers is you might have one or two network adapters or four in a server. So you have all these virtualized networks, whether it be Nuage or NSX, and they tunnel down to just a couple of MAC addresses. Our network has thousands of paths because we're photonic. So we can, we're affected like a multi-core network underneath these overlays. That is very different. In fact, your original question is very different than anyone else in the market. And the catalyst, what's the analog to multi-core for you? Is it convergence? Is it just all that complexity? It's diversity, we would say diversity. So that we don't have just one path or two or three or four paths. We have thousands of paths in our network between any two points. It's as if, you know, if you were to travel from Boston to New York, there's probably three or four ways you always go. You take 90, 86, you know the drill. Like the Mary. But there are probably thousands of ways to get there. Not just the Mary. What if you could know that? And what if you could dynamically change your path? Like you do in some of the new phone apps. Ways for networking. Ways for networking, yeah, yeah. Right? And that is the nature of what we're doing, which is very different than anyone else. So how would you think about network management differently? I mean, what does network management look like in the scenario that you're building? So really what we want, our ultimate manifestation is we want to be invisible. We want the network to be invisible. We are so well integrated into VMware and OpenStackNow, this is part of our announcement last week, that the bulk of network administration goes away. It just goes away. And we look inside of VMware environment, there are five types of network traffic in a VMware environment. There's the inter-node traffic, there's the storage traffic, there's the logging traffic, there's the control traffic, and there's the VMotion traffic. There's five different network types. We can distinguish every one of them and create a network topology that's optimized for VMware. Period, that's what we do. So it is intended to be idiot-proof? Really, the network manager? We're trying to... Does any of our network manager go away? It never really goes away. I mean, we saw the same thing years ago. So I was the exec sponsor for the Iceland acquisition. And so at that time, at EMC, I was running in mid-range, so the biggest part of the NAS group was my group. But we realized there was limits in our architecture and that we needed to get to a scale-out net, a scale-out, that's a four-gain slip, a scale-out file system, an Iceland represented that. And at the time, everybody was very worried, like, what happens to administrators? And I remember we used to measure this. We used to say, in the VNX era, administrator could manage 100 terabytes. Well, that same administrator in Iceland can manage dozens of petabytes now. Do you think they're unhappy? Because you know what, we took the misery away. We took the misery away. And we want to do the same thing on the network. I want to talk about scale-out because so many network functions have been historically scale-up. You have to add on top of the infrastructure, the bigger switches, bigger paths. I mean, my understanding was that there were functions of networks that just couldn't be scaled out. Are you saying that's not the case? Yeah, that's not the case. So we have customers using us in the data center. We have customers that are using us around metropolitan areas, and we have one customer that's part of this announcement, Perseus, who deployed a 34,000-mile network around the Pacific Ocean. I can't tell you who the unused customers, they'd kill me, but you'd be amazed who the customer is, okay? And they control all of that from New York City. They provision VLANs between the West Coast of California and Tokyo. By telling the network, I want a VLAN between here and here. They don't touch dozens or hundreds of devices in the network. And they control it all from New York City. They've told us now they could never build such a network and have so few people managing it. We fully distributed the control function so we can do that in the data center around a metropolitan area and a global scale, and we do it with exactly the same product. There's not a difference between our product used in the data center around a metropolitan area or at 34,000 miles. It's the same product. What we require is a wire. We need a wire. So undersea cables is a wire to us. It has different characteristics. It took us some time to get that to work. It was a little scary for a while because we had never tested a network that big because we're a startup, but we got it to work. And so we've proven that you can scale these things. Let's talk about the route to market. You did a deal with Arrow. One of the least talked about things in the Dell EMC acquisition is the channel. That's right. And Gartner has their swim lane model, which is pretty good modeling. You got the box movers. You got the sort of guys doing more integration, focused on Oracle, SAP, VMware, moving up the stack. And then you got this sort of new breed of cloud guys doing DevOps. How do you see that channel shifting and how are you trying to take advantage of that? That's a great question. So this is probably the thing we're most excited about. I mean, this announcement to us is not just a product announcement. It's a year of learning of our battleground market. So as I said, we shipped our enterprise-ready product in Q4 of last year, and it's the basis of our revenue. But we've learned a lot. And startups need to be very adept and just keep learning and refining or selling motion. And what we have learned is that there's a new customer out there. And we call them the cloud builders, or cloud architects. And they think very, very differently about the infrastructure. They think about it not from the Cisco switch or the EMC array. They really think about the applications and the business problem they're trying to solve. They think more like what we would call a platform to system integrate, more from the top down. And because the nature of the customer is changing, there's a new partner that's emerging. And they're cloud builder partners. And a lot of the reason why we spend a bunch of time with the arrow guys and recruited them and have been very focused on 100% channel oriented business model is to create this next generation partner, to enable them. And what those next generation partners need are technology tools that they can use as cloud builders to shape the infrastructure the way they need it. Because I think gone are the days when people want to buy a Cisco switch or an EMC array and that has a fixed function, it's more DevOps-y as you pointed out, right? And these guys, they don't want very rigid components that can't be assembled in unique ways or they can't change. So Plexi, one of the things that's unique about Plexi is we're really a tool for the cloud builders, whether they be a partner or an end user. And that tool can be shaped by the mechanisms we talked about a minute ago. So your value prop for them is you're eliminating a lot of that non-differentiated stuff down low and allowing them to focus further up on their value delivery. For example, like this 34,000 mile network, it has no MPLS in the whole entire network. There's none. I don't know anybody likes MPLS, okay? There's none. So all that complexity is removed, all right? Talk about, you referred to this announcement a few times, you should tell people what that is. It's the Cloud Builder 2.2 software suite and the new Switch 3 100 gig E hardware platform just announced last week. Tell us what's breakthrough about these. Yes, so again, the focus on this Cloud Builder customer and partner is fundamental. It's kind of our biggest learning actually for this year. And so the announcement was about a set of capabilities mostly in software and then one additional new hardware platform that we announced. The capabilities are really in two dimensions on the software side. We announced Plexi Connect, a new set of plugins and software that allow us to look into OpenStack and VMware environments to understand the nature of the requirements in these environments and then present that information to what we call Plexi Connect. And as a set of new services that I won't go through them all but I'll hit one or two. For example, one of our customers wants us to isolate specific lambdas in our network in a topology around multiple switches so they can basically snap in Docker or NSX or other services that want to partition the applications but then have a physical control all the way through the network that says, this blue goes all the way through all of these switches end to end and no traffic shall be on this or not. In some respects, it's the opposite of the algorithms we've been building up to this point which shape the network dynamically for them. This is isolating these capabilities but these are tools that these cloud builders want to achieve their business objectives in this case, security and isolation end to end. Also, we put in capabilities where things like VSAN or ScaleIO or other storage oriented traffic would get distinguished in nature of that traffic and separate out whether it be metadata traffic for say VSAN or regeneration traffic of rebuilds or user traffic and separate those out on our networks seamlessly. So the end user, if I want to challenge what we have is you can't demo it because it just works. The network is dynamically shaped. We're so well integrated into VMware now that when you do a VMotion, we actually effectively wire a separate network for the period of time you're doing a VMotion and then return that resource. So in a lot of enterprises, they shut off VMotion. They say you can't use that. Well, why is that? Because I built this leaf and spine, I put all these applications on this virtualized environment and say VMware, I have hundreds of apps and then I have this VMotion which blasts terabytes of storage east-west through my network. You have nondeterministic results. You don't have that on our network. We just see the VMotion. We say it's going from here, let's create a separate network for that duration and the VMotion is done, we return the resource. How important is the hardware component to your business? Is there an NFV sort of future for you? So, see, I call me old fashioned, right? But I think people want to buy like whole products in the enterprise. I think you have the big web-scale dudes who have engineering groups bigger than EMCs and they can build whatever they want. So I don't spend a long time trying to sell to them. Then you have some of the big banks that have a lot of resources. I think people want finished products. I think they want hardware software because I think they want to know that it's going to be serviced. One of the powers of EMC and one of the greatest lessons I learned at EMC is the importance of the customers knowing that you will make them right, that you will take responsibility for the entire stack and that's what we do. So hardware is important from that perspective. A lot of our innovation isn't in hardware. If you look at our hardware, it's totally off the shelf. I mean, we use Broadcom chips, we use Intel processors and we use DWDM components that are used to fiber to your home. There's no innovation there. We put it together in kind of an interesting way but this announcement is about actually a white box that is a Tomahawk based system and we manifest our software on that. It's table stakes to a solution. Yeah, it's a table. You want a solution. Some people will disaggregate you. I think that's a mistake. I don't think enterprises really generally want to do that. Let's talk about scale because that's really your big challenge. Are you scaling the company? Your recent round, you raised 35 million. I was just out in Palo Alto last week. Of course, I go out there all the time. If you walk down the Rosewood hotel and you see a million people that you know. We're East Coast guys. And now we see the big East Coast company is merging with an Austin based company. Still going to be here, a big presence here, but so talk about scale. How you scale this company? Have you raised enough money? What's the plan? So we talked about the three things, right? Raise money, stay alive, sell stuff, build stuff, sell stuff. So I always do three. I'm always raising money. It's just what you do, right? So you're always raising the next round. So I'm always raising money. The good news is there's lots of capital out there. So I don't really focus on that as my day job, because I think if we sell enough stuff, then all the other problems will be solved. So I focus on selling. I ask a lot of executives who have startups, particularly storage guys, like you've seen the Pure IPO, you're seeing a bunch of other storage companies, flash guys, can they reach escape velocity the same way that Isilon did, and Threepar did, and Incompellent to a certain extent did as well. What does it take today to reach escape velocity? So I think this is a change, a time of huge change, right? So some of the traditional classic OEM models, I mean, I can remember several startups ago, you had to have an OEM component in the storage business, otherwise you couldn't really take off. And so that's very confused now, because all the big guys are very confused. They're all restructuring it somewhere. The good news is that there's no accident that we signed Arrow, because they were the emergence of a new kind of partner that's reaching these customers. And the good news is that the big guys are pretty confused, and they're everyone, whether it be IBM, or HP, or Oracle, or EMC, they're all concentrating their selling motion at their top customers. And that's just a huge opportunity for us guys that are coming into the market. Sure, around that play. You look at Aristide, Aristide plays a little bit different. They're placed to concentrate in Cisco's top account, and they're doing a great job. They're doing great, Jayshree's unbelievable. She's unbelievable. They're doing a great job. We're a different cat, right? Because we're different, right? We're a different cat. I mean, you talk about Cisco, obviously it's going to be impacted by what you do. Is Cisco, do you see them as being a competitor of yours? Are they a potential acquirer? Are they a partner? Can they get to do what you do? Well, I mean, take three years and, you know, $75 million, $85 million, and you can build anything, right? And it's just that easy. Having built a lot of things in my career. So, yeah, anybody could do it. And it's hard. I mean, it seems a bit complicated. And, you know, having lived in the enterprise for over 30 years, it really has to work. It's not, I'm sorry, no offense West Coast, it's not a website. So it has to work. And that's a real challenge. Cisco is, you know, depending on whose math you use, 80% market share in the enterprise, they're everywhere. Is that a handicap or is that an asset or a handicap in trying to restructure? It depends. Sometimes people are looking for alternatives. So we're very favorable on that. Generally, if it's just a port refresh deal with Cisco, I don't even bother to show up. There's no reason for me to show up. They should buy from Cisco or they should buy from Arista. We're looking at the cloud builder. Okay, so what if it's not a port refresh deal, but the customer is still nervous, small company. What's the customer imperative? How do you get sort of get them through that not whole? So, we, I can't tell you my wholesale strategy. I can't do that. I can't really answer that again, because I wouldn't play my cards. But we look for where the agents have changed. We look for, and they could be the networking guy. They're generally a cloud builder though. They're generally not, they're not married to ECMP or MPLS or any kind of protocol thing. They're not religious about those things. They think more about the infrastructure and what they're trying to accomplish. And we know we have a customer, and this happens in every meeting, that ends up transforming into a customer. They go, ah, this is different. You tell me how to solve this problem. Because I've been trying to solve it for the last 20 years, and I'm really not happy. When they have the epiphany like, this is different. These, there are different ways to solve these problems that I've been working on for a long time. And you've removed so much complexity from this problem, I need to think differently about it. That's when we have them. That's when they're on. So it's those guys that get, that it's a 10x value proposition with virtually infinite ROI that they couldn't do it before, and now they can do it. And it's not about capex. So I think a lot of what's happening in a networking space is everybody's so focused on capex. That's a joke. I don't know why people even bother. Because people don't have a capex problem because 90% of their spend is OPEX. So people would trade OPEX for capex any day if you give them a case. And one of these use cases that we haven't really gone into, one of our customers, they're taking their head counts and shifting them from the networking to other things. Like thousands to hundreds. And the TCO associated with that is off the charts. You touched on it before. I mean, Amazon, they don't need it, but the enterprise is going to spend money on a solution to save time any day and reallocate that time at the telephone numbers that far exceed the capex of a box, no question. All right, Rich, awesome as usual seeing you. Thanks so much for coming in. I'll give you the last word. What do you want people to leave this session with around Plexi? If you're a cloud builder partner, we're looking for you. Great. That's key to our go-to-market. That is so key to our go-to-market. And what type of people are you hiring right now? Mostly go-to-market aside. So guys that sell into the enterprise and that know how to work with partners to keep beating that drum. And then a few kind of fill-in engineering guys. But we generally have a good beat on those because we know the talent base. Well, Rich, thanks for coming into the studios today. Really appreciate it. Good to see you again. All right, thanks for watching everybody. We'll see you next time. This is Paul Gillan with Dave Vellante. Thanks for watching.