 Live from Orlando, Florida, it's theCUBE, covering .conf18, brought to you by Splunk. Welcome back to Orlando, everybody. Of course, home of Disney World, I'm Dave Vellante with Stu Miniman. We're here covering Splunk's conf18, hashtag conf, sorry, hashtag SplunkConf18. I've been fumbling that all weeks too. Maybe by day two I'll have it down. But this is theCUBE, the leader in live tech coverage. Phillip Adams is here. He's the CTO and lead architect for the National Ignition Facility. Thanks for coming on. Thanks for having me. Super interesting off-camera conversation. You guys are basically responsible for keeping the country's nuclear arsenal functional and secure. And effective. Yeah. And effective. So talk about your mission and your role. So the mission of the National Ignition Facility is to provide data to scientists of how matter behaves under high pressures and high temperatures. And so what we do is basically take 192 laser beams of the world's largest laser in a facility about the size of three football fields and run that through into a target the size of a BB that's filled with deuterium and tritium. And that implosion that we get, we have diagnostics around that facility that collect what's going on for that experiment and that data goes off to the scientists. Wow. Okay. So and what do they do with it? They, they model it. They, I mean, that's real data. So they use it to model real world nuclear stores. Sometime back, if you actually look on Google Earth and you look over Nevada, you'll see a lot of craters in the desert and we aren't able to do underground nuclear testing anymore. So this replaces that and it allows us to be able to capture by having a small burning plasma in a lab. You can either simulate what happens when you detonate a nuclear warhead. You can find out what happens from a, if you're an astrophysicist understand what happens from the birth of a star to full supernova. You can understand what happens to materials as they get subjected to, you know, 100 million degrees. For real? For real. So, well now, so some countries North Korea in particular up until recently was, was still doing underground testing. Are you able to, I don't know, in some way shape or form monitor that or maybe there's intelligence that you can't talk about but do you, do you learn from those or do you already know what's going on there because you've been through it decades ago? There are groups of the lab that know things about things, but I'm not at liberty to talk about that. I love that answer. Okay, so maybe you could talk a little bit about the importance of data. So, you know, your group's part of Lawrence Livermore Labs. I've loved geeking out in my career to talk to your team, really smart people, you know, some sizable budgets and, you know, build, you know, supercomputers and the like. So, you know, how important is data and, you know, how's the role of data been changing the last few years? So, data is very critical to what we do. That whole facility is designed about getting data out. And there are two aspects of data for us. There's data that goes to the scientists and there's data about the facility itself. And it's just amazing the tremendous amount of information that we collect about the facility and trying to keep that facility running. And we have a whole, just a line out the door and around the corner of scientists trying to get time on the laser. And so, the last thing IT wants to be is the reason why they can't get their experiment off. Some of these experimentalists are waiting up to like three, four years to get their chance to run their experiment, which could be the basis of their scientific career that they're studying for that. And so, with a facility that large 66,000 control points, you can consider it 66,000 IoT points. That's a lot of data. And it's amazing some days that it all works. So, by being able to collect all of that information into a central place, we can figure out which devices are starting to misbehave, which needs servicing, and make sure that the environment is functional as well as reproducible for the next experiment. Yeah, well, you're a case in point. When you talk about 66,000 devices, I can't have somebody going manually checking everything. Just the power of IoT, is there predictive things that let you know if something's going to break? How do you do things like break fix? So, we collect a lot of data about those end point devices. We have been collecting them and looking at that data in the Splunk and plotting that over time, all the way from like capacitors to motor movements and robot behavior that is going on in the facility. So, you can then start getting trends for what average looks like, and when things start deviating from Norman, set a crew of technicians that will go in there on our maintenance days to be able to replace components. Philip, what are you architecting? Is it the data model, kind of the ingest, the analyze, the dissemination, the infrastructure, the collaboration platform, all of the above? Maybe you could take us inside. I am the infrastructure architect, the lead infrastructure architect. So, I have other architects that work with me for database, network, sysadmin, et cetera. Okay, and then so the data presumably informs what the infrastructure needs to look like, right? I.e., where the data is, it is centralized, decentralized, how much is it, et cetera. Is that fair, assertion? I would say the machine defines what the architecture needs to look like. The business processes change for that in terms of like, well, how do you protect and secure a SCADA environment, for example, and then for the nuances of trying to keep a machine like that, continually running and separated and segregated as need be. Is what? As need be. Maybe, yeah. What are the technical challenges of doing that? Definitely, you know, one challenge is that the Department of Energy never really shares data to the public. And for, you know, it's not like NASA where you take a picture and you say, here you go, right? And so when you get sensitive information, it's a way of being able to dissect that out and say, okay, well, now we've got a user community of folks that now want to come in remotely, take their data and go. So we want to make sure we do that in a secure manner. And also that protects scientists that are working on a particular experiment from another scientist working on their experiment. You know, we want to be able to keep swim lanes, you know, very separated and segregated. Then you get into just, you know, all of these different components, IT. The general IT environment likes to age out things every five years. But our project is, you know, looking at things on a scale of 30 years. So, you know, the challenges we deal with on a regular basis, for example, are protocols getting decommissioned. And not all the time because a, you know, the protocol change doesn't mean that you want to spend that money to redesign that IOT device anymore. Especially when you might have a warehouse full of them in the backup, man. So, and obviously you're trying to provide access to those who have the right to see it. Like you say, swim lanes, get data to the scientist, but you also have a lot of bad guys who would love to get their hands on that data. So, how do you use, I presume you use Splunk, at least in part in a security context, is that right? Yeah, we have a pretty sharp cybersecurity team that's always looking at the perimeter and making sure that we're doing the right things because, you know, those of us that are builders and they're those that want to destroy that house of cards. So, you know, we're doing everything we can to make sure that we're keeping the nation's information safe and secure. So, what's the culture like there? I mean, do you got to be like a PhD to work there? You know, to have like 15 degrees, CS expert. I mean, what's it like? Is it diverse environment? Describe it to us. It is a very diverse environment. You've got PhDs working with engineers, working with, you know, IT people, working with software developers. I mean, it takes an army to make in a machine like this work. And, you know, it takes a rigid schedule, a lot of discipline. But also, you know, I mean, everybody's involved in making the mission happen. They believe in it strongly. It's, you know, for myself, I've been there 15 years. Some folks have been there working at the lab 35 years plus. So, all right. So, you're a Splunk customer, but what brings you to .conf? You know, what do you look to get out of this? Is this, have you been to these before? Yes, you know, so at .conf, you know, I really enjoy the interactions with other folks that have similar issues and missions that we do and learning what they have been doing in order to address those challenges. In addition, staying very close with technology, figuring out how we can leverage the latest and greatest items in our environment is what's going to make us not only successful, but a great payoff for the American taxpayer. So, we heard from Doug Merritt this morning that data is messy and that what you want to be able to do is be able to organize the data when you need to. Is that how you guys are looking at this? Is your data messy? Are you kind of, you know, this idea of schema on read? And, you know, what was life like? And you may or may not know this. It's kind of before Splunk and after Splunk. Before Splunk, you know, we spent a lot of time in traditional data warehousing. You know, we spent a lot of time trying to figure out what content we wanted to go after. ETL and put those, put that data sets into rows and tables. And that took a lot of time. If there was a change that needed to happen or data that wasn't onboarded, you couldn't get the answer that you needed. And so it took a long time to actually deliver an answer about what's going on in the environment. And today, you know, one of the things that resonated with me is that we are putting data in now, throwing it in, getting it into an index, and, you know, almost at the speed of thought than being able to say, okay, even though I didn't properly onboard that data item, I can do that now, I can grab that, and now I can deliver the answer. Am I correct that, I mean, we talked to a lot of practitioners, they'll tell you that they're, when you go back a few years, their EDW, they would say was like a snake swallowing a basketball. They were trying to get it to do things that it really just wasn't designed to do. So they would chase Intel, every time Intel came up with a new chip, they'd say we need that, because we're a star for horsepower. At the same time, big data practitioners would tell you, we didn't throw out our EDW. You know, it has its uses, but it's the right tool for the right job, horses for courses, as they say. Is that fair assessment? That is exactly where we're in. We're in very much a hybrid mode to where we're doing both. One thing I wanted to bring up is that the message before was always that the log data was unstructured content. And I think Splunk turned that idea on its head and basically said there is structure in log data. There is no such thing as unstructured content. And because we're able to rise that information up from all these devices in our facility and take relational data and marry that together through DB Connect, for example, it really changed the game for us and really allowed us to gain a lot more information and insight from our systems. When they talked about the enhancements coming out in 7.2, they talked about scale, performance, and manageability, you've got quite a bit of scale and I'm sure performance is pretty important. How Splunk doing, what are you looking for them to enhance their environment down the road? Maybe some of the things they talked about in the Splunk next that would make your job easier? One of the things I was really looking forward to that I see that the signs are there for is being able to roll off buckets into the cloud. So the concept of being able to use S3 is great, great news for us. Another thing we'd like to be able to do is store longer live data sets in our environment and longer time series data sets and also annotate a little bit more so that a scientist that sees a certain feature in there can annotate what that feature meant. So that when you have to go through the process of actually doing a machine learning algorithm or trying to train a data set, you know what data set you're trying to look for or what that pattern looks like. Why the S3 is signed with you, just because you need a simple object store with a get put kind of model and S3 is sort of a de facto standard, is that right? Pretty much, yeah, that and also, if there was a path to, let's say, Glacier, so all the frozen buckets have a place to go because again, you never know how deep, how long back you'll have to go for a data set to really start looking for a trend and that would be key. So are you using Glacier or not? Not very much right now. There are certain areas, my counterparts are using AWS quite a bit. So Lawrence Livermore has a pretty big Splunk implementation out on AWS right now. Okay, cool. All right, well, Phillip, thank you so much for coming on theCUBE and sharing your knowledge and last thoughts on Conf 18. Things you're learning, things you're excited about, anything you can talk about. This is a great place to meet folks, to network, to also learn different techniques in order to do data analysis and it's been great to just be in this community. All right, well, thanks again for coming on. Appreciate it. Thank you. All right, keep it right there, everybody. Stu and I will be back with our next guest. We're at Orlando Day One of Splunk's Conf 18. You're watching theCUBE.