 Live from San Francisco, California, it's theCUBE at VMworld 2014, brought to you by VMware, Cisco, EMC, HP, and Nutanix. Now here are your hosts, Dave Vellante and Stu Miniman. Welcome back to San Francisco, everybody. We're here at VMworld 2014 or at Moscone South, just inside the lobby to the right, come by, say hello, love to see you. So this is the fifth year for theCUBE at VMworld, and we've seen the transformation of VMware from what used to be called the software mainframe. The marketing guys got a hold of that term and have really evolved it into the software-defined data center. And Stu and I are really excited to have Bindi Gillon. He is the chief architect at Nutanix, hot company, one that we introduced you to quite some time ago, and we've been following the ascendancy of Nutanix. Bindi, welcome to theCUBE, good to have you. Thank you, say hi. So a lot of talk about convergence, hyperconvergence, software-defined, you guys got ahead of that whole curve. We saw some announcements today, everybody's trying to parse through those and understand them. Give us the update on Nutanix and then we'll get into the sort of architecture and the technical piece because you were essentially at the de facto CTO of the company, so please. I mean, we have come a long way since we started. I mean, early on we were talking about converging the hardware, so the storage hardware and the server hardware and why converging them makes sense. There are many people who thought that may not work because the compute side and the storage side have different refresh cycles and all of that. So we have gone through all of that now the entire world understands why convergence is a good thing, why server side flash is a good thing. Now we're talking more about web scale. Once you actually take monolithic San and NAS boxes and the capabilities and sprinkle that capability into hundreds of servers, what you have to now grapple with is the scale at which the storage functions have to be developed. And that's the web scale movement that we are at the forefront of. Yes, Stu and I have been tracking and folks at Wikibon, the whole hyperscale Stu piece bleeding into the enterprise. It's something that you've covered extensively, you've written about it. Obviously you've written about server sand. So it's happening. Yeah, Dave, when I joined Wikibon, and I came from a heavy enterprise focus and very different architectural design from the web scale guys. So really exciting when you get to talk to James Hamilton from Amazon, talk to Facebook, what they're doing with open compute. And while Nutanix actually takes some hours from the competition because they say, oh, this web scale thing is nothing but a bunch of marketing. But Binyam, I'm wondering if you can share with the folks, your team, I've talked to guys in your organization that are former Facebook and Google architects, worked on some pretty amazing technologies that those really big companies and are helping bring those to the Nutanix portfolio. Right, I mean, so what we have done is instead of tackling just the storage problem at the get go, what we have said is, let's take a pause and see how have the web scale companies evolved the data center. So if you look at what they did, they did not start off with a NetApp or an EMC or NASA sandbox, what they said is that they need massive scale. And to do that fundamentally, first they did is they changed the storage architecture. So that means you converse the storage piece and the compute piece. The way they did that though is they shipped the compute piece to where the storage lied. And that is the invention of big data, that's the invent of MySQL, that's the invention of MapReduce and all of that. All that goodness does not directly apply to enterprise data center because there's a lot of legacy applications that need POSIX interface, a lot of applications that still need the notion of a virtual machine. So we took those web scale concepts and said, first we will build a web scale foundation on which now you have strong consistency. That's why we had to delve into multipack source that was distributed. We had to build MapReduce from ground up that would look at your metadata and then give you a lot of the features of DDU, Snapshot and all that. And then we said, we have this web scale foundation that will get rid of silos in your data center and now we build storage on top. And that's what we have done with the employees that have come from these various backgrounds that bring that mindset and the religion of web scale to us. So, Bini, the other thing I want to let you address is those really big companies, a lot of times they're talking about a single custom application or maybe a handful of applications, most of the enterprise is usually, I mean, dozens or hundreds of applications. So why is the architecture still applicable when I get away from a single app to go to virtualization and kind of the typical enterprise environment? So yes, so it's not about the architecture, it's about the discipline. So I'll tell you what, if you want to do one thing right and get web scale is follow the scale out principle. The scale out principle says that if you have a service running on a given node, that service should use resources that are proportional to the size of the node and not proportional to the size of the cluster. So as you're increasing the size of your cluster, the services should not become hotspots and bottlenecks. And that fundamental religion is what is important. Now you say, okay, I understand that, how do I keep metadata? And suddenly your answer would be that you need NoSQL, you need MapReduce. And so those are the things that we learn from the big guys, not necessarily exactly their application stack and how they build it, right? And then we say, enterprise is a complex environment, right? And great companies like VMware have created a lot of solutions around that. So how do you take the goodness of VMware, the goodness of Microsoft and KVM ecosystem and also going forward into the cloud and provide web scale to those environments in a seamless way? You know, when you put Nutanix in a box, it just works. You don't have to do anything special on the high provider. So that's the key. It's the ability to scale out seamlessly. You say seamlessly, everybody. When I hear seamlessly, I think, oh, I can add capacity and not take any downtime and not take any application disruption but there are architectures. I won't name them right now in the queue, but I might later, where the customers have said, well, yeah, it's scale out, but when I scale out, I take some downtime or I take disruption or I have to go through hoops. I presume that's not the case with Nutanix but I want to push on that a little bit. I mean, philosophically, that to me is not really seamless scale out. What's your take on that? I mean, again, if you look at Google, if they add a new node, Google doesn't take a downtime. I mean, those are fundamental things that it's a no-brainer to make sure that you enable that. In fact, we focus a lot on a single click upgrade for an entire cluster, no matter how large it is. Upgrading not just our software stack but also upgrading things in our ecosystem, like the hypervisor, the device firmware and stuff like that in the same way. And that is, you know, the fundamental tenet around web scale is making sure that there are no silos because when you have broad convergence into the data center, what you have to understand that is you don't have a single monolith that everybody can point to and get to the data. Now you have scattered that all over your data center and you better be sure you're not creating silos around that because the biggest pain point in the data center is not about just adding nodes necessarily, it's about agility, right? So I have a virtual machine running here, an application, I want to move it from one place in the data center to the other. Well, in VMware, you can just click on the VM and do vMotion, you're done. Well, not really. What if this virtual machine had 10 terabytes of data sitting here and now that's sitting in a different silo? And silos could be performance silos, capability silos, hardware silos. Protection silos. Yeah, protection silos, all sorts of silos. And then you say, okay, the data is sitting there, my application needs to go to the other end of the data center. What do I need to do? You need to implement fine-grained data migration, right? And that's where the complexity of a metadata layer, that's built on NoSQL, just like Cassandra, that enables us to do it on a fine-grained level. The hard-working set follows the virtual machine, not just within a data center, but also when the VM moves or the application moves to a different data center, doing deduplication, compression, at a fine-grained level across data centers and clouds, that dissolves the silos in the data center. Yeah, we've done studies at Wikibon that quantified the cost of migration, an example, the cost of migrating a storage array well north of $50,000, 20 plus percent of the original cost of the array. So based on what you just said, I'm presuming that you're able to take that cost down dramatically. Yeah, so if you're running on the Nutanix virtual computing platform, and you're moving things around, you don't have to think twice. There is no cost, there is no planning you need to do it. Just do it. Just do it. Do what you want to do. Yeah, I mean, Dave, when we did the server-sand architecture, you know, David Floyer said, taking out that migration cost is one of the key tenants of server-sand, and of course, Nutanix is a great example of that kind of architect. Benny, you talked about the cloud a couple of times. When I first talked to Nutanix, you guys were VMware, it was the environment that you worked on. You guys announced something, I believe it's called CloudConnect, relatively recently. Can you talk a little bit about that, and how do you extend from kind of the data center beyond the individual data center to the cloud? And what did you guys talk to me for that? It comes back to solving silos, right? So now today we have a divide between the public cloud and the private cloud that is growing. What the enterprise admin wants to do is I want to move applications from my private cloud to the public cloud. I want to burst my applications into the cloud space, and also some amount of freedom of choice. You know, they want to go to Amazon, they want to go to Azure, and stuff like that. So what we have done is, you know, taken our architecture, our software, and put it on EC2. So now we have our controller virtual machine running on EC2 and providing the same services as it provides in the private data center. Again, the reason we can do it is because we don't have a hardware crutch, and it runs beautifully on Amazon as well. So now, currently, we put the cold and old data onto Amazon, but going forward, it's about moving workloads from one location to the other, and all seamlessly. Yeah, so, I mean, at this show, we've seen some of the challenges that VMware is having with competition. You know, it's a multi-hypervisor world in many ways, Microsoft's been gaining sure Red Hat's doing well in their spaces, Amazon and Azure, of course, in the public cloud. And then there's the big question mark, things like Docker and containers. From kind of an architectural standpoint, and what you're looking at, how is Nutanix positioned for multi-cloud, multi-hypervisor, VM container bare metal world? Are you guys a VM only, or how does your architecture span? So, I mean, if you look at our intellectual property, it's all in software, right? Now it's a question of how do you package that software? I mean, you can package it in terms of a virtual machine, you can package it in a bunch of processes, running in a C-group on bare metal. I mean, nothing stops us from taking our software and run it on a laptop, for example, or in a container in a docker on bare metal. What we want to do is focus on the enterprise pain points, and if VMware is at running in a lot of servers and our customers are running VMware, so that's where we provide our value. And now we have seen the shift from VMware to Hyper-V and the need for open source KVM. So we have provided our functionality there and there's no reason we can't go to Docker as well. All right, so Vinny, I know we're running short on time here. You came from IBM research. I mean, IBM got over a hundred years of creating some of the best innovations in the IT industry. I mean, way too many to name. I worked for Lucent for a bunch of years back in the 90s. Bell Labs is the one I always greatly respected being an engineer at training. What did you learn working for research that you've brought to Nutanix? How do you guys kind of look at the future and look at the research that you guys are doing? Yeah, I mean, so one of the main things to learn from research is you need to empower engineers, empower technical leaders with the ability to fail, right? So the permission to fail is the recipe to succeed, right? So fundamentally, you have to understand that there are hundreds of startups out there that want to build a Nutanix killer now, right? Now they have the permission to fail. Most of them would fail and some of them would succeed. How do you bring that culture into the company as a first class citizen and say everybody is allowed to take big bets, take risks, and make it work as if it's a research organization. And that's what keeps me up in night and make sure that that culture is what we maintain as we are growing, doubling every year. So I wonder, Bini, if we could, maybe sort of one last touch point. How should we think about the Evo rail? Everybody's talking about, oh, it's kind of going after Nutanix and others, so we had Sam from Delon this morning. He's like, no, no, no, no. Very complimentary. Nutanix is like IO killer. They're unbelievable at IO and we're positioning. So how should we, what's Nutanix's point of view on this rail announcement? So it's a very good thing for us, actually. It's a great validation for what we have been talking to our customers. We've been telling customers, look, you need a control plane that is different from the control plane you're used to. A lot of the things around analytics about VM management need a different look at it. So we have built Nutanix Prism UI earlier this year. We announced VM analytics. So we look at the health, CPU, memory, network, storage, layer for virtual machines. All of that now becomes easy for us to actually tell the customers, look, you've got to be open to a new control plane. And by the way, this control plane is going to be the same web scale as you're used to on the data plane from Nutanix. Awesome, hyperscale coming into the NEC prize. You guys are making it happen. Vinny Gill, thanks very much for coming on for theCUBE. Appreciate it. All right, keep it right there. We'll be back with our next segment right after this. This is theCUBE, we're live from VMworld 2014.