 Lie from Palo Alto in the heart of Silicon Valley. It's theCUBE, covering IO. Brought to you by IO. Now, here are your hosts, John Furrier and Peter Burns. Okay, welcome back everyone. We are here live in Silicon Valley. This is theCUBE on Sand Hill Road at the Rosewood Hotel for IO Conversations part of the data center as a platform conference here and our next guest is Mark Musselman from Vapor IO and Bruce Taylor, Data Center Dynamics. Guys, let's jump into it. You guys are leading a panel on a panel around strategic co-location, data center in the future. What is the big trend that you guys are seeing right now? Because the data center we all know is people want to get out of that business. But ultimately, what does it look like? What does it look like going forward? At least customers want to get out of the business. Customers want to get out of the business of managing data centers. Sure, they do want to get out of business of managing data centers. But there's so many options now, right? Do they want their own co-location? Do they want to create a private cloud that's managed by someone else at some level? Maybe it's just infrastructure as a service, but it is not in their facility. Then they have all the public clouds. I think it's gonna get, it'll be more complex for a while and then it'll simplify as the software catches up and as the technology and networking catches up. Complex, I mean, you're gonna have your own data center for a while. You're gonna have Amazon, you're gonna have Azure, you're gonna have software, you're gonna have any number of providers that accumulate to give you the best experience and bang for the buck. They call that hybrid cloud right now. That's kind of what everyone's kind of talking about. They are. Yeah, Microsoft has something called Azure Stack, which is basically run Azure by yourself and then also burst out to their public cloud. But they're not there yet. They will be later in the year. We're waiting for that to see what proprietary books they have into the stack, but that's the whole other conversation. That's a concern, right? But that's the complexity part of it. To simplify it, you have to be able to be unmanned. So if you're gonna have all these data centers spread out, you don't have your own people there and you probably don't have the provider's people there, especially when it gets more and more distributed for things like IoT and where you wanna get out closer to the edge, where the sensors are and where the customers are. So you have to come up with telemetry that lets you instrument as well as monitor and maintain all of these applications that are gonna get more and more distributed. So actually take the software that used to run the data centers and then move that out into a distributed set of data centers. That's the complexity that's trying to be automated. That's right. Okay. So the dynamics and impact to the customers would be? Well, I think one of the things that we just began to touch on today in the panel is that where strategic decisions are now made corporately has changed radically. It used to be that the CIO and the CTO were kind of the holders of strategic IT. And what's happened now is that's moved way down the stack so that people down at the development level, at the DevOps level, and at the systems admin level and architecture level are actually making strategic decisions because it's workload-centric. It's application-centric. It's no longer a huge, monolithic, physical billion-dollar decision you're making about your data center. You're making these micro-decisions all the time, right? And that's only gonna increase. And it will increase to the point where it's actually all done in software. Let's build on that for a second. So that in fact, there are people who won Nobel Prizes for this analysis. In fact, generally speaking, we institutionalized the work around the assets, right? Right. That's kind of, it's not particularly cars. Yeah, exactly. So we have an asset and we figure out who has claims on that asset. We do have to share it. Okay, we're gonna institutionalize the work to make sure everybody gets their claims. When the asset is a fungible, heavy-duty, cat-backed thing that leads to one institutional form. Now, the CIO was the king of those assets. But now, as data become recognized as an asset, and the question is how does it get shared? How does it get utilized? How does it get deployed? It means that because it's so shareable and it's not necessarily limited to one group of person, that those decisions naturally start to move down in the organization to where the data is being used. Right. Is that kind of the underlying dynamic here? Right. Absolutely. I think that that's the, you know, the CIO, Gardner calls it what, the bi-modal CIO, bi-modal IT, I call them bipolar. What's really happened is that every enterprise is somewhere on this continuum. Some are way up the scale, you know, in hyper-scale data centers, very purpose-built, very vertical in nature, very granular in how capacity decisions are made. Right. It's down not just at the rack level, but the half-rack level and on. And yet there are enterprises, and I shouldn't say this, but I think of Google versus Century Tire, right? Their needs are entirely and utterly different. And my guess is that Century Tire is still thinking about their data centers as depreciable assets, 15-year term, right? Major investment where the rest of us are thinking, how granular can we get this? How what's the smallest increment I can possibly buy in order to run this app? I think you're right about how it was the CTO and CIO, and now it's kind of further down where it starts. You have a systems architect that knows that he needs to deliver a real-time analytics application, right, and so he's got to figure out where to run Kafka, Cassandra, Spark, and HDFS in order to satisfy a need for the business. And where he picks that, the CIO and CTO don't really care, and certainly no one above them. They just know they have to create this new application. It's probably microservices container-based. So where can you do that? Does he go and buy it as a service? Does he go to Google and get TensorFlow as a service? Or does he run TensorFlow open-source himself in a colo? You figured that out, but just be on time. Well, so let me try to make that a little bit explicit, you can tell me I got it right, Mike. When the execution of the decision can move down the stack, the need for governance becomes much more crystal clear. When we had an asset, and we were making decisions about that asset, the governance decision was embedded in the asset. Now we're talking about distributing decision-making out, and it becomes that much more incumbent upon the business to offer very clear, streamlined governance models. And that seems to be one of the drivers behind this continuum of data center, because the decision about how to source your technology needs is very, very closely tied to your overall governance model, the governance model you should use, because are you gonna do it through contracts, or are you gonna do it through normal, HR coercive kind of methods, but you gotta start doing it. And one of the reasons why people are using third parties is because it forces that governance conversation because you're embedding it in a contract. At least I think so. Do you see that as well? Yeah, we definitely see that. You need the SLA, you need the governance, and you need a way to turn up and turn down, I think is the big thing. And you need to buy in, you need to buy one piece of gum. You don't need to buy the Costco case of gum, but you also have to have all of those things in place. Well, it's fantastic to have a developer, so the Chinoff group, Jim from the Chinoff group was talking about this earlier about how the business conversation on security drive down. So let's go to the DevOps guy, the guy who was racking and stacking servers, worried about top of rack switches versus this and the other things. So they had there, when they call them pets and cattle, it's an analogy, whatever it is. But that guy now has to have invisible infrastructure. He's composing solutions. So that need is to actually access a fabric of resource and independent of what's underneath it. Deprivision, compute, storage, networking, to run a specific workload. That's the discrete unit I think you were referring to. So that has to have everything embedded. Is that kind of where that? Exactly, that's to build on your point for a second. Governance is gonna happen in software. It's not gonna happen by heavy handed HR and policy emails. It's gonna be embedded, it's gonna be part of AI, it's gonna be part of everything that we do. The RFP will go away, the SLA will go away. And the sooner the better, frankly. But certainly within a five year time horizon, 90% of all capacity decisions are gonna be embedded in rules in software. You know, I'll push you on that a little bit. I'm not, no, no, no, I'm going to, no, no. But I'm not gonna disagree with you, I'm gonna push you on it. So that, so for example, many years ago, I sat down with the CEO of a large pharmaceutical company who was making a decision about whether or not to buy SAP. This was many years ago. And he ultimately decided to buy it and his exact quote was, I'm sick and tired of process arguments. So by buying SAP, he poured SAP on his business and that became the process model for his business. The similar type of thing is gonna happen here. But where we can use software from an administrative management function or standpoint to start providing or pouring some of this governance over the organization. But there is still a lot of very subtle things we have to think about when we think about data as an asset and how the business uses data and how the business treats data. And those have to be a little bit more explicit than just bringing into software just kind of spread peanut butter across the organization. No question, I agree with that 100%. However, when SAP lost the battle was in 2007, right after that cost really became the driver. And if cost is a driver, doing it, automating it in software is the answer. Well now cost is one driver. Now revenue generating, this is where the agile thing comes in is interesting. So cost was a 2007 problem or indicator of where you were. Now it's agility. So now it's the, how do I drive revenue from my business? That's interesting around how you make a DevOps infrastructure as code. So the question to you guys is, for this new cloud trend, we were having this debate in our office last weekend on our show, our Cube show for Friday, Cube's day on our podcast. Will every net new application, whether it's net new for the enterprise, net new meaning not legacy and all cloud native apps eventually run in a cloud? Yes. Absolutely. Without a doubt. Yeah, that seems like a no brainer. So that's, I think everyone's kind of realizing, okay, but not necessarily Amazon per se. Amazon for some stuff, but the new cloud could be this kind of co-location, facilities outsourcing, hybrid cloud or ultimately off-prem in a way. Well, there could be, but not 100% off-prem. But I'm trying to get to is, if every application new net application goes to some sort of cloud model, what does that look like and what's the impact of that? Well, there could be, you could see a kind of a fragmentation sort of in cloud and hosting where there's very specific providers that are AAA rated in a certain area. Maybe someone only focuses on, we were talking about AI and then we were also talking about, you know, medical MRI sort of latency that you need. Maybe someone's better for that than Amazon. In other words, it's not gonna be defined by the way the hardware is deployed. It's defined by the way the stack is deployed. For service. Absolutely. And maybe someone is the best at hosting machine learning applications and ML and GPU. And maybe someone's has a great cryptography solution that people wanna, you know, host there, but it's all gonna be public. It's all gonna be cloud. Well, look, with hooks to something. Well, the business side. This industry's on an inexorable march to public infrastructure. It always has been. Because when you go to public infrastructure, digital technology has shown that the larger your market, the more you're gonna make money. And by having it look more public, more and more and more, you increase the number of applications that you have access to. So we're trying to take out with some as asset specificities, get to public infrastructure, increase the size of our markets. And that is unassailable. We're gonna get there eventually. It's multi-tenancy squared. I mean, that's the way they're gonna have to do this. Even if you're a niche guy coming up, right? Or a Colo provider trying to transform into more infrastructure as a service, it's all, you have to harness multi-tenancy. And that's where software-defined networking comes in. But even two years ago, you'd have a conversation with the CIO about the different options. And it was, do you wanna own the hardware? Do you wanna Colo? Take the hardware, blah, blah, blah. It all centered on the hardware. It was, the market was segmented by where the hardware was. And you've just introduced, you're amplifying this notion that it's not where the hardware is. It's the quality of the services that end up being provided. Where a medical service in the cloud is gonna have a different security, processing location, administrative cost, everything else, compliance, and you're gonna buy that assortment of services. That's right. Great, guys, great stuff. Last minute, I just want you guys to just talk about the impact of the customer. If you're an enterprise buyer and looking at their data center, what should they be thinking about right now? What are the things that they should be doing? What steps do they need to take to be safe, secure, and on the right path? Well, I think there's more options than ever. And the proliferation of open source has only expanded that. So if you look at where we're headed, even with things like OCP, Open Compute Project, Open 19 that LinkedIn is driving, there's a commoditization, there's a push towards open source, and there's more options than ever. And so I would make sure that you're on top of all of those options and understanding. You don't have to buy proprietary software for everything anymore. You don't have to buy proprietary hardware anymore. You probably can get everything on a lease, right? Then to not do that could be a detrimental effect on your business, yeah. Great. There's any comments ahead of that? I think agility drives a great deal of this, right? The requirement now is on 90 day cycles, not on 10 year cycles, right? So the more that people understand that and begin to move in that direction, one of the things that I often say is if you don't now have a DevOps team in your organization or contracted to you, you're dead and you just haven't gotten the memo yet. Okay, man walking in the data center. That's a good show format, including Netflix, next to Game of Thrones and House of Cards. So guys, thanks so much for sharing the insights here in theCUBE. We're here live for the IO Conversations, live at Silicon Valley for a special CUBE presentation for data centers and service event and we'll be back with more after this short break.