 Live from the Moscone Convention Center in San Francisco, California, it's The Cube at Oracle OpenWorld 2014, brought to you by headline sponsor Cisco Systems with support from NetApp. And now here are your hosts, John Furrier and Jeff Frick. Okay, welcome back, and we're here live in San Francisco for Oracle OpenWorld 2014. This is The Cube, Silicon Angles flagship program. We go out to the events to extract the symptom noise. I'm John Furrier, the co-founder of Silicon Angle Media. Join my co-host, Jeff Frick, with The Cube. And we are here with Tim Frazier, the CIO of the National Ignition Facility with Lawrence Livermore Labs. You guys doing some really awesome stuff. Welcome to The Cube. Thanks, and pleasure to be here. We love to talk to CEOs, entrepreneurs, also people in the trenches putting technology to use. You guys are doing a lot of material science research, cutting edge, using a lot of compute, storing a lot of stuff. So the first question is, what's the environment like for you today versus, say, ten years ago? Paint the picture of what it's like now in terms of access to tech, your workers' environment, your colleagues. What did you say just ten years ago? Right, you know, I think the most disruptive change came in probably five years ago. We, like a lot of our colleagues out in the industry, took advantage of virtualization, both in servers and also in storage, to really create an infinitely scalable cloud that can be used by scientists, both at our location, which is about 40 miles from San Francisco, and then also with scientific partners around the world. So you guys are virtually live in a VM environment with Oracle. How did you wrap storage around, obviously, NetApp's sponsor of this segment? What are you guys using NetApp for? And how does that fit into the whole virtualized and Oracle piece? Is there like software written? Is it Tire-in? What's that? Right, all work together. Yeah, that's a great question. We use NetApp in two ways. We use it as fiber attached storage for the boot disks for all of our hypervisors. We have about 40 of those. And we also use it as network attached storage for all of the scientific data that our instruments generate and that ultimately get analyzed and visualized and shared around the world by our scientific partners. So is that raw data or data exhaust, as they say? So that comes in often off the technology that you guys are watching. Is this a stored and then you guys apply algorithms to it? How does that work? Think of it. Yeah, that's right. You know, in science, we have one rule, which is never ever rewrite history. You know, you're okay to revise it by making new calculations or writing new algorithms, but the raw data that were generated by the experiment are stored forever. So we are the traditional write once, read many archive. The algorithms generate anywhere from 2X to 10X reduced data based on the raw data that the cameras and the other instruments saw. You know, Jeff and I were talking just before you came on about we had the TED event in San Francisco last week with IBM. And one of the things that came out of us is all about reimagining the world. So share with us some of the discoveries you guys have done with the data. I mean, because this is the classic case of, do you store all the data? Do you not? Having that raw data is certainly great. But then using the data to make a discovery. Could you share some just anecdotally what you guys are doing and some discoveries you've made? Right, well, you know, what I'd like to do is tell you a little bit of a story. You know, the big idea for the science that we do was had in 1960. So we had a young researcher at the lab who heard about the invention of the laser, which is another great California story, Malibu, California. And he knew that that was the pathway to deliver enough energy to a hydrogen target to get it to fuse. In essence, creating a tiny miniature sun on Earth. Now over the years, we've used very, very large compute clusters to simulate those events. And we've built a number of laser systems out of the lab to experimentally test that those simulations were on the right track. Now we've finally arrived at a facility that has a large enough laser to actually deliver on the promise that that researcher had. So over the years, you know, the data have been archived and analyzed and re-analyzed to understand how to perfect, you know, our equations and our theories to actually achieve something that humankind has been after for 50 years. So you didn't rewrite history, you recasted the history. Well, we recalculate what we think the significance of history is. To get it to a baseline today. That's right. Any other progress from there? Is there like prototypes coming out? Is this going to spawn potentially new ways to do fusion or whatnot? We surprise ourselves every day. So we've made fundamental breakthroughs in fusion science, fundamental breakthroughs in material science. We found a phase shift in carbon that nobody knew existed. We've modeled the interior cores of all the planets in the solar system. So we have a pretty good idea of what the big ball of iron at the center of the earth actually looks like and behaves like. And we've also done fundamental breakthrough science on how black holes form, how supernova form and how planets form. But Tim, talk a little bit about how real fundamental core science that you guys are doing at a really deep level. How does that impact people's day-to-day lives? Where does some of those discoveries kind of come out and add value? Right. Yeah, there's certainly people better qualified than I am to speak to this. But in a lot of ways, the science that we do is based on hydrodynamics, which is the study of how fluids flow. So in terms of the experiments that we do, we have confirmed the astrophysics models of how planets form. So we now have a better, more scientifically-based picture of how our solar system may have come together. Some of the spin-off benefits from the simulations have been in how fluids flow through artificial hearts. So if you think about a blood vessel traveling through a heart that isn't made of human tissue, a long-standing problem has been that the materials inside the heart caused the blood vessels to rupture. So by better simulating and having more high-fidelity simulations of how the fluid flow, we can make better hearts and save lives. So how long have you been at Lawrence Livermore? Well, I've worked at Livermore since 2007. But I've spent my entire adult life inside or outside of one or more of our national labs. So I guess you could say my entire adult career. So talk about how things have changed with the, obviously Moore's Law has been working on the microprocessor forever, but really the change in terms of cloud and big data and kind of storage and IO really developing, especially via virtualization. And some of the scale and things that you guys do in this core science that you couldn't imagine so many years ago. Right, yeah. So one of our fundamental stressors, if you will, is that as the instrumentation that we use to capture the data increases in resolution, meaning we go from a 2X to a 4X to an 8X, let's say a megapixel camera, right? That implies drastic increases in the amount of storage that's required to ultimately archive the data and then give a representation to the scientist. So had we not been able to economically scale up storage and along with that compute by virtualization, we would have held back the progress of science. So for us, we think of it as being a game changer. And are your hypotheses keeping up with the amount of horsepower that you can now leverage? We could always use more. So what's the big thing about NetApp that allows you to do this? And you guys have a cloud. What's some of the landscape components of your technology? Like you got virtualization, big part of that. So using compute clouds and on-premise compute or using, and then them as the raw start for the data and how are they playing into the data usage for you guys? Right, well, so I would say the first and most important aspect of the NetApp technology is it just works, okay? So reliability and availability of our scientific instrumentation and the uptime of our facility is, sort of where we begin and end our day in terms of what we think about. So it being reliable and performant and being able to take that somewhat for granted is its largest contribution. Being able to economically add capacity or add and failover and redundant recovery sites using some of the features said that they haven't on tap have also been critical to how we think about data management and data storage. And then as we scale out our compute layer, we've had to add IOPS and other capacities to the filer heads just to make sure that we could keep up with the rate of computing. It's like the old Star Trek, you know, Scott even need more power in the engineering room. Everything comes to Star Trek reference, Jeff, on the key, we always know that. To bully go where no person is coming. You guys are doing that there, but is there an IOPS, Scotty, I need more IOPS kind of thing going on with storage? And it was that more of a cloud issue. So when you talk about IOPS, give us an example of where you were stuck in a jam and you just saw them with more IOPS. Yeah, right. So, you know, another story might actually help. If you imagine taking a thousand computers, you know, that all have their own internal disk drive and virtualizing them and placing those disk drives now into a, you know, that filer head. Essentially what you've done is you've focused a very random IO pattern onto, you know, a single set of devices. So the issue hasn't so much been a need for more IOPS. It's been the ability to effectively deal with random IO patterns without pressure. Let's talk about the big trend with the cloud. Is that like with scale out commodity hardware, the need for more data, big data intensive or fast data or data intensive applications are being put on the system. So do you move the data to the compute, compute to the data? Do you have a philosophy on this whole debate? Well, you always move data, you always move to compute to data, you know, at scale. Right, it's the only option that works. And in some cases, our data are dominated by time series, right? So if you think about the stock market, the price of the stock is a function of time. Sort of our largest data by volume looks like that. And we would not be able to do it if, you know, we were moving the data to the compute. Give us a taste of what would the consequences be if you didn't have it? Well, you know, you'd months? Well, you'd have a very smart person waiting on an answer to come back who would give up and walk away. And so the discovery or the innovation. The creativity gets. Would not happen because, you know, of a very human reaction to having to wait a long time. Yeah, yeah. And having that data also waiting for that, but also there's other parts of the project and other people waiting too for kind of that answer. It's like a relay race. Yeah, that's absolutely the way we think about it. You know, no one person has the answer. It takes teams of experts collaborating over months sometimes to actually form a clear picture in their mind about what's happening. So any delay or disruption to how that, you know, normal human activity unfolds is a killer. So often the big data world is described that you know that you have enough data, you throw it into the data lake, you're going to start to discover insights, right? Versus having a hypothesis that you're going to go in and test a hypothesis and see what you can get back according to that hypothesis. Are you seeing much of the former where, you know, you've got the systems that can start to deliver some insight and information prior to you really having a hypothesis to go test? Yeah, this is, is there a magic box? You know, if I just feed the magic box with enough, you know, data then somehow data mining or machine learning algorithms will find the answer. You know, we haven't done a lot of that. I think, you know, the variety and the volume of our data is such that, you know, we always start with a hypothesis and then use the data and machine learning algorithms to confirm that hypothesis. But, you know, the person that taught us how to do that was Galileo. It's called the scientific method. Right. And, you know, we've stayed true to that. Yeah, and there's always a human aspect to the data. With this God box mentality has gone, certainly with big data, we've talked to folks all the time who say, hey, often the question that you want to ask is like the third, fourth one or question way downstream. You just want to start playing with the data and that idea of machine solving it is moving more towards, they add to the value to the user, these are experience. Or to help you move down a hypothesis chain quicker. Yeah, more visualization. You know, our God box are the scientists. So, facilities by some God boxes. Exactly. Sons of God. That's right. Our mantra is, you know, feed their head. Yeah. Well, that's great. I want to ask, what's the coolest thing that you, outside of the fusion work you've done? A weird quirky little cool thing that you've seen enabled by technology in the past couple of years that just caught you off guard. You're like, oh my God, it's another moment. Well, certainly my nest at home, you know, and all that that implies when you really think about, you know, what that will mean to, you know, a smart home, but also the Raspberry Pi. You know, I have three Raspberry Pies that I play with at home with my children and, you know, just the capacity to, you know, automate and distribute computing just floors me. You know, that really brings up a good point that we talk about Jeff all the time and also with Dave Vellante, our co-host, is that to make our culture back, you know, this idea of hacking has come back. I went to the 30th anniversary of the Mac in Cupertino last year and all the old timers there arguing, I had to put 1K software. I took it from MacPay for the MacDraw and they were arguing about who's going to get that 1K of RAM, but that old hacker kind of homebrew mentality is now back in full swing, drones, Raspberry Pi, mobile. First Lego League is a wonderful example of that. Amazing trend that really doesn't get a lot of attention, but that's the new generation. I mean, you're seeing that young hacker open compute in the data center. I mean, you're seeing guys making their own data center to their own servers, so that's a cool trend. So I want to ask you, with that in mind, you agree with that trend if so, what do you think the young generations coming in, the new blood, the guys coming out of school, never loaded Linux on a machine, who probably don't even know what a DBA is to be frank, they're coming into the workforce, whether scientists or workers. What's the mindset of some of the young talent coming in, what's your observation? I think they want it to just work. Certainly, the expectation of our smartphones, nobody wants to hack their smartphone, they just want it to work. They want it to be infinitely fast, infinitely deep. They've forgotten about file systems, right? There is no hierarchical storage anymore, it's all cloud-based. And I think that's what the industry really has to think about over the next couple of years, we're going to be pushed in so many ways to forget about things that we hold dear, that you and I grew up assuming would always be there, a file system, there is no file system on an iPhone, there is no file system on a Google Drive, even though they make it look like there's one, yet somehow you can find all the information you need. So that brings up the question, so that's the user expectation, the preferred user experience, so you guys are pushing the envelope, I'd say great success at NetApp, as large-scale engineering continues, if you take your mission forward, you're looking at even more larger scale requirements. What are some of the things that you look for as this new large, large scale, I guess extra large scale, architecture rolls out, I mean it still has to work, so that implies engineering's getting done. It has to work, it has to be something that you can understand. I mean the basics haven't fundamentally changed, dividing, conquer, reduce complex problems to something that you can actually understand and explain to your mother. It's sort of a reasonable expectation to have, but for us it's all about requirements. Understand what the user wants, and by user in this case, I mean the scientists and the physicists, understand what they want, and always fight complexity. Tim, we talked a little bit before you went on air about kind of the different points of view of a mathematician and a statistician and a traditional scientist versus a computer scientist. Really everything is confidence levels and probabilities, and there is no right or wrong, and computer scientists are always looking for the one and the zero as much as you can. Talk about how those worlds are kind of clashing around big data, and how kind of the expectations of what can be delivered and or the bounds by which the confidence of an answer should be described, kind of what you're seeing there. Right, so the best example I can think of that illustrates I think what you're talking about is actually climate. If you think about all the instruments that we have out in the ocean and all of the environmental samples that we're taking in the atmosphere, we're taking them over great distances, right? So if I want to build a model of, let's say how the climate in the Atlantic is behaving right now, I'm not getting that at centimeter resolution, I'm getting it let's say at a meter or 10 meter resolution, which means there's a lot of uncertainty baked into what's going on between the measurements you're taking. So when you take all of that data, and we're talking hundreds of petabytes now, and you plug that back into models that try to make predictions about what the climate will do next, you inherently have great uncertainties in how you're calculating right now. The computer scientist writes a piece of deterministic code that does the calculation from time stuff, one to time step two, but the airbars involved in the measurements are so great that what you end up with is a complete fiction. It doesn't match reality except when you take another guy's model and you mesh it in with your model, the air tends to cancel. So what we're finding is multiple approaches with different types of calculations actually help with this problem. To blend and blending those together. Right, exactly, it's sort of a merge cancel. Interesting. And you can obviously understand why if we had better climate models that could predict snowfall or rainfall or even wind over longer time periods, it would make huge differences for how humankind will evolve. Tim, really appreciate you coming on theCUBE. I know you're busy, CIO of a great organization. What do you think, just to kind of end the segment, just kind of throw that, throw the projection forward. Outside of your scope of work where you get to discovering new things every day, what in society do you see kind of being the low-hanging fruit for the use of technology, big data? Is it climate change? Is there other societal benefits that you see that we could possibly get to within our lifetime that we would want to have happen? I would say a couple of things. Drug treatments and human patients, understanding how new drugs interact with the body, certainly the spread of disease and the delivery of energy to the developing world are sort of the challenges of our time. Those are big ones. Tim Frazier here inside theCUBE. Always getting the expert opinion. Great to have you on. Great content. This is theCUBE inside the Cisco booth. Great to hear the NetApp success story as well. Great shout out to NetApp. Thanks for supporting theCUBE. Really appreciate NetApp's ongoing support of theCUBE and SiliconANG. Really appreciate it. We're here live in San Francisco. We'll be right back with our next guest, Walton Walcott, Oracle Open World 2014. It's all about the data. It's all under the covers. It's all about the applications in the cloud. We'll be right back with theCUBE right after this short break.