 Live from the Julia Morgan ballroom in San Francisco, extracting the signal from the noise, it's theCUBE, covering Structure 2015. Now your host, George Gilbert. And we're back. This is George Gilbert. We're at Structure 2015. And we have a very special guest, Stacey Higginbossen, who along with a very select few colleagues powered the rise of GigaOM and also the Structure series of conferences. And she's here to tell us what she's seen today and how it's changed over time. Well, welcome, Stacey. Thank you, George. Yes, I had the privilege of being at the inaugural, I guess this is the second inaugural Structure conference, but at the initial Structure conference all the way back in 2008, it was. And that was a much smaller event. And I remember being there on a panel with Vijay Gill who has moved around so much, but we were talking about, my God, what is a Paz? And at the time I had no idea. So it was a perfect panel for me to moderate. And back then we were like, oh my gosh, what is this cloud thing? Amazon had launched AWS two years ago. We were still like, we couldn't even sell this event. And now we're, it's huge. And we didn't know, you know, were we gonna move to the public cloud? And the consensus was no one's gonna put any workloads in the cloud if they're a real legit business. From today where we've got Disney putting things in the cloud, we've got the federal government having things in the cloud. But there's actually kind of been this pushback where people are like, okay, we accept that there's the cloud, but now we're like gonna build our own kind of thing and we're gonna take some of the lessons of the cloud with DevOps and the agility that kind of the cloud offers and we're gonna bring it back home. So there's a lot more kind of talk about what we've learned and taking it on-prem. So yeah, let's dive into that a little bit because for the longest time we were told virtualization, carving up these underutilized servers was gonna sort of give everyone the benefits of the public cloud and, you know, the lower CAPX but the operational processes that made the public cloud so efficient didn't quite migrate to the private cloud. When did we first start seeing that and when did we become aware that that was a challenge? I think some companies are actually still dealing with that. Even today we had someone from GE and he was talking about the cloud really isn't about OPEX for us, it's about agility. And I was kind of shocked to see that because that felt like something that we had been talking about four or five years ago. So four or five years ago. So I was like, oh, meanwhile we still have like, you know, Facebook and Google and these hyperskill guys and they feel like they're moving, you know, they're talking about microservices and utilization rates that are far above and away like what any of these private cloud guys could even be achieving right now. And they're like, I never want anyone to think about a VM anymore, you know, I just want to let people push services live no matter what. Well, this is an interesting comment. Do you think that, you know, there's been a law of sort of legacy systems that's held for like six decades which is they run, you don't touch them. Could we see that sort of all new green field apps get built on this new hardware infrastructure with this new, with these new software processes and we sort of ring fence the legacy systems? Is that a? I think so. So one of the things I talked with Diane Bryant yesterday morning and one of the things I thought was great about her talk was she, she spoke about all of these new IOT kind of options out there. So like John Deere building these new connected tractors and agricultural, I'm going to say tractors as a service kind of, just like farms as a service. And all of that's being built on their own on-premise cloud things. And so all of these new services. On Deers. On Deers on-premise clouds. Yes, like on-premise agricultural field, clouds, words. So all of that's being built on their own cloud and a lot of other companies like GE are building their own IOT services on their own clouds. And so all of these are new greenfields opportunities and they're being built on-prem, on clouds that they're hosting. And so there's going to be these legacy systems that no one's going to deal with. I mean, IBM's still selling their mainframes, right? Then there's going to be these like fun testing and QA and weird workloads that you're like, send them to Amazon, send them, I don't know, maybe to Google. Since, you know, maybe Google's going to get into the enterprise a little more seriously or Microsoft. And then there's going to be the things that, you know, you just leave on-prem because they work. And maybe those won't even be on a cloud. So it's interesting that you mentioned the Internet of Things apps as greenfield apps because we don't have anything that they're replacing but to think of them as that their first or natural home might be on the vendor's premises, on the enterprise's premises. Have you gotten any insight as to why that might be? Because of their data, some of that is just so precious to the company that has it or maybe it's like a hospital or medical data or like, sorry, from a medical device. So like, if you're Johnson and Johnson and you're selling some sort of connected medical device, you want to keep that data secure as possible because my God, if something goes awry and you're like insulin pumps start shooting out sugar into somebody's like, I don't know what you shoot insulin into. Well, like your body and your glycemic index goes off. Yes, that's terrible. That's bad. You really want to lock that down. So that's one example. And the other is because, you know, this is precious. One day you might be able to sell it to somebody else and you don't want that in Amazon if you can help it. So what about the concept of data gravity or someone told us the other day, data mass? Like data isn't drawing the planet. Have you been talking to Dave McCrory? Yeah. Yes, we give him credit. But the notion that, you know, if you've got information coming in from all sorts of devices everywhere, it might more easily come into data centers that are publicly managed because they're more strategically placed and that they have the cabling, the optical cabling for low latency, you know, replication or messaging if need be. And so it almost, that infrastructure might lend itself to Internet of Things apps. I have no idea how to respond to that because I still am a big fan of like the problems of bandwidth and we haven't gotten around that. But a lot of the data that we're talking about with the Internet of Things is smaller. It's time series, it's, you send it someplace, you do the processing there, you handle it and you've got a, I don't know what this is called, this thing that's happening with my fist. But you've got a decision that's made and then you send it out. Right. So with that, the concept of data gravity doesn't really necessarily have to apply as much. And I just don't think, like in an ideal world, that would be the case. But in a, the way the business models are evolving, it is appearing to be very much like I keep my data, I'll let you have access to some of it and through some sort of kind of data exchange. The challenge is we still don't have those business models yet. So I think we're kind of at a place where we're like, maybe we have a magical data lake, but everyone I'm talking to in like the IoT world is like the problem with data lakes is nobody can grab anything out of it when they want it in time. It's not curated, yeah. So they hate that. And that's what I think of when we think of kind of data gravity. And the problem with data exchanges is nobody has a business model yet that makes it really work. So it's really kind of this data mess. And so I don't, I don't really know the answer. Okay, so that's interesting. The takeaway is that we're kind of in between generations, between sort of distributed applications and the new era of internet of things applications. Well, those are, I mean, I still think those are distributed. Yes, but they are characterized by capturing and processing much more data towards the origin of that data. Right, at the edge. At the edge. So what are some of the things that surprised you about what you saw here or maybe perhaps disappointed you that you didn't see more of? I would have loved to talk more about this, but. We have a conference. I was like, we have a conference hopefully in June that we're going to do a little bit of this at because this is something I'm really excited about and kind of figuring out architectures for this exact problem because it's data synchronization, like how do you do with that? It's got bandwidth problems. It's figuring out like, how do you move software and processing like around at the edge? Because sometimes you're going to want it at the edge. Sometimes you're going to want it at the core and it may differ. So would the sweet spot of the applications that we're dealing with now and the infrastructure be the software defined data center? The sort of terminology that we were driving towards eight years ago. I don't think so. So the software defined data? Not for IoT, but for structure. So for structure, I don't know. I mean, so the software defined data center, I feel like maybe the data center is everywhere. I feel like that's probably a better way to start thinking about it. That's interesting. The infrastructure is everywhere and how do you build for that? Okay, so in other words, the construct of four walls and a bunch of guard dogs, you know, that's a little too physical. Now, what we've got is sort of something virtualized and the distance between the elements is what we don't have to worry about. It's spreading, so. Okay, that's interesting. But I'm just kind of a crazy person out here, so who knows? Well, that's why we want to talk to you. Not the crazy part, but the vision part. The out there part. Maybe the two go together. All right. So if you look back a couple years and then take that as a connect the dots and look out a couple years, what do you think at Structure we'll be talking about a couple years in the future? Oh man, this is the kind of question I ask people and I'm always waiting for their answer and I never want to think about it on my own. I think security, we've got to figure out. I think machine learning is going to be a big topic. I'm really interested to think about what kind of hardware, how we kind of build data centers around that. I've been waiting for years for ARM to be invading the data center. I'm still waiting, so I don't know how we're gonna, what we're gonna see on that front, so. And I'd kind of like someone to take Intel down a peg just in terms of the monolithic processor. They've got 96% of the market in servers. That's a lot. I heard a statistic that was kind of interesting that if you, once we get to the 10 nanometer geometry for transistor size, all the transistors that we will need for all the data centers in the world will be produced in one fab, or the capacity of one fab. And the rest is memory. Any thoughts on what that might mean for what the data centers and Intel look like? That's a crazy stat. Is that just for servers? Basically, yes. Okay, because, I mean, we'll still need chips for everything else, so. But less and less Intel chips. Right, well that's why Intel's gotten into the fabrication business, like fabbing other people's chips. So I would imagine we'll see more aggression from them in making other people's chips because that does make sense. So that would be my thoughts on that is, ooh, they better start signing up some more customers. Have you thought about systems management and as we disassociate the physical boundaries of a data center from sort of the capabilities of a data center, what does it mean to manage that infrastructure? Oh. That wasn't meant to stump you. No, no, I'm like, oh, George. I don't think a lot about systems management outside of IoT, so I do think about it in terms of managing sensors. Yeah, so, okay. So I think about like, God, ILM on sensors and deploying any sort of. ILM being the information lifecycle management. But not really information lifecycle management, but not technically that. But when you do updates on sensors, how can you roll that back? Because right now there is no management for if you have a massive array of sensors in a factory, for example. If you update all of them and it breaks something, you really can't control that right now. And so getting some sort of orchestration layer on top of that is really important. So I think about those kind of things, but I don't think about it really for the data center right now. Okay. All right, Stacey, as always, fun to talk to you and have you on, very insightful. We'll have to leave it there. This is George Gilbert, and we are at the Julia Morgan Ballroom in downtown San Francisco at the Structure 15 conference, and we'll be back in a moment.