 Live from the campus of MIT in Cambridge, Massachusetts, it's theCUBE, covering the MIT Chief Data Officer and the Information Quality Symposium. Now, here are your hosts, Stu Miniman and Paul Gillin. And we're back. This is theCUBE at the MIT Chief Data Officer and Information Quality Conference in Cambridge, Massachusetts. We're here for two days live streaming with a lot of speakers from the program, a lot of federal government and just government in general speakers on this program, wrestling with really big data issues, big scope issues, but also those with big payoff. And it doesn't get much bigger than the Department of Defense. Our guest here right now is Kevin Garrison. He's the Principal Director and Director of Analytics, Deputy Chief Information Officer for Business Process and Systems Review. That's sort of a typical DOD title, isn't it? It's about that long. But what it really comes down to is analytics. You're doing a lot of analytics projects at the DOD trying to figure out areas of cost savings. And analytics is such a dynamic area right now. It seems like there's a new Apache project every two months. There's something new out there to absorb. And federal agencies dealing with, of course, large bureaucracies often, change is not internalized quickly, typically. How are you keeping up with all this change? So it's, to be honest, it's a challenge, right? I mentioned in the panel we had this morning, change in DOD is like turning the enterprise into the wind with a canoe paddle, all right? It takes a long time. So while, and we have a number of impediments we kind of put in our way or just grew up over the years to get stuff. So our procurement processes are somewhat drawn out. There, for example, was a new cool analytical tool I wanted. It would take me about a year to do all the necessary actions to get the stuff procured and then approved to run on our network so I could apply the new shiny object tool to the top of our data. So that means that you got to pick a shot, at least that's how I interpret it. I can't afford to go, don't have a lot of money, so I can't go, let's buy five and see if one of them works. I got to kind of find one that's new but not bleeding edge, do enough market research to make sure that there are people who have used it enough to understand if it will scale for DoD. Because DoD is not just a big company. It's not a larger Walmart or a Shell Oil or GE Capital. It's all those and more, it's really an economy. We have like 10 times the amount of warehouse space of Walmart. We have a huge medical enterprise. We fight wars. It's a country. Kind of our mission thing. And so the business part of the department is really languished in terms of management attention because the mission is to fight war. So we've been fighting wars on handheld devices since the turn of the century, right? I don't have wireless in a Pentagon where I work. I think next week we might. Sometime real soon now. Well, you're going to need it for Pokemon Go. Well, we can't play at work with productivity go. So we have that problem. And then we also have a challenge of, because of our procurement processes, some of our hardware and software is out of date. So I personally use a computer that has five years old laptop, four megabytes of RAM. We're running an office 2010. I got 2016 at the house, but 2010 at work. So it's hard to wheel a new toys in and run on that set of infrastructure. So this is a challenge. So we're behind. So I have to, for my effort, I have to kind of that I'm doing on behalf of the department. I have to kind of pick not bleeding edge. I've got to pick something that's somewhat proven that we're confident can get the job done, even if it's not the best thing that's available in the world right now. Enroll with that. We're still in an effort, about 18 months maybe going on 24 into doing this analytics. We've assembled probably about 60% of the data we need that we'll eventually have in Data Lake, whatever you want to call it. Data reservoir is a term people like better because it's built by design as opposed to a freak of nature. So we're about to crawl walk one stage. We're about getting to walk. So we'll get there sometime that towards the end of the calendar year we'll be able to have the skill sets on the field and the tools on the field to start doing some of the stuff that the other people here have been doing for years. So I'm curious to talk about kind of the budgetary constraints and trying things out. So how does open source fit into your choice? My personal opinion is there's a procurement law about how the department. My personal opinion is the only affordable solution for analytics going forward for any large scale enterprise is you really got to look hard at open source. Otherwise, you're vendor bound, proprietary, switching platforms like switching ERPs, not for the faint of heart and not cheap. So all the reading I've done, all the work I've done, all the people I've talked to just feels like open source is delivering, the community is delivering the solutions. The scale that are vibrant, that are fresh, that are making the best use quickly of the latest technology and thought. And so if you're in the analytics business, you really want to be looking hard at moving up. As somebody pointed out yesterday, you've got to go from where you're at right now. Whereas Russell said, you go to war with the army you have, not the one you want. So you may want to be open source, but you're stuck with a bunch of legacy infrastructure. And you just got to understand that, work your way through it. But my personal opinion is open sources, where are things, where it's headed. And we've been looking hard at open source tools. Because the licensing costs kill you. I mean, the old school, buy a license per seat, per core, whatever, when you try and scale that across a DOD enterprise, it just gets to be dollars B. And you just can't afford, you can't sell, I should spend dollars B on licenses to do data analytics. It's just not a salable business case. You're in charge of finding cost savings through analytics. And as taxpayers, we all appreciate that. Can you point to any specific examples? So we did, at the deputy secretary about two years ago now, chartered an effort to go review the part of the department that wasn't the army native in the Air Force. As he said, the army name Air Force had taken huge budget cuts, reducing cutting divisions out of the army and ships out of the Navy. But the back office part of the department, which we call the Fourth Estate, that was coined some time in the past. It's all our functional agencies that do logistics or finance and accounting or information systems and the OSD staff itself that works for the secretary. I hadn't really taken any cuts. We'd grown as a part of going to war, so the staff had gotten bigger and the agencies had gotten bigger, but now that the war was kind of winding down, we hadn't taken any cuts. So we formed this organization I'm in to help do the analytics. It involved also organizational visits to go identify where it was spending perhaps not necessary and could be reduced. And we did that, found about $6 billion worth of savings, whether we'll get them all. And finding them and getting them were two different things, right? But we identified, how do we put it, $6 billion of savings opportunities? And then the implied or secondary task was to go get a data environment stood up so you could see if those savings ever happened. So that's what I've been doing for about the last 18 months is trying to make the data part on the back end. Can we go get the data to show was there a reduction in spend and was it in the areas where we asked, we thought it would happen, kind of what happens? And that's what I've been doing. So we think we found about $6 billion. I'll probably know or we'll know sometime next year how much of that we think we're really going to get over. Because we had, the $6 billion was over a five year timeframe. So you got to have a couple of data points to draw a 10 trend line. So we have end of 15, I'll have end of FY 16 here probably around the turn of the calendar year. And then we'll see if we have a trend line going downward and at what kind of slope and what do we think happens next year? How do analytics play into finding these cost savings? So you might not believe this, but it's very difficult to know how many people work in Organization X. That stuff is data buried in siloed systems that and some of them give you two different answers to the same question. So stitching together, how many people work at Organization X and how much do they cost is a non-trivial problem. You have active duty military, you have reserve military that are on active duty. You have reserve military who could be civilians who then drill as a service. You have contractors, which are very, very hard to track because it's based on how much money you have. So adding up how much money does it cost to have, does it cost to have a DOD CIO, one of the organizations I'm in, all right, was a bit of a drill. I mean a real knuckle drill and it's like spreadsheet hell and then you got to augment that with data of other things. So wait a minute, the data doesn't have Joe in there. I know Joe works here. He's right down the hall to kind of get an answer. And the DOD says an organization about 120 to 140 bodies that are government, military, civilian and then about another couple, 300 contractors. The contractor workforce fluctuates, right? So the question is how many people do you have in the DOD say how much do they cost depends on in what day are you talking about? Do you mean over the course of a fiscal year, what was the average cost? So the precision about what are you asking? So we can't do that, but we could say it's bigger than a bread box, smaller than a house, go with the median and then make some adjustments. Can we twin 10% off the top or 5% off the top? It was kind of how we went about it. The other part is what are you spending? So the spend goes out the door on bodies and on contracts. So we were able to go get data sources and mine contracts data because previously we generally focused on what you budgeted, but a budget's kind of a fiction you tell your boss to get them to give you the money you need to get done what needs to get done and then stuff happens. So then what do you actually go spend it on after you may differ from what you said? So we've been working real hard on what's the money actually being spent on, which is hard, but we've been able to do that. In that you can find, for example, we found we pay 100 different prices for a software product. Some are good, some are bad. Somebody got a good deal, somebody got a bad deal. So we've seen that, right? Now can you get all the people who are writing bad deals to come over and write a good deal that's a little bit harder to do. We have in some cases, we've negotiated large enterprise license agreements for software products, Oracle, Microsoft or whatever, and we have a contract vehicle that has pretty good pricing on it. But we can run the numbers and show there's somewhere between 30 and 40% that we can find that people are not using that contract vehicle to buy software. In many cases they're paying, they didn't get a good deal. Subcases, they got a better deal. So we've been able to turn that into savings. We found some savings in mobile because we weren't leveraging our buying power. Every little people, 50 people were buying mobile licenses instead of having to 500 buy a license to get better pricing. And then we weren't managing it. So we've got pool managers to manage minutes. If you don't use your phone, we take it away from you. That turns, when you scale across the area, that ends up to be real money pretty quick. So we've had a number of those kinds of things. We've done a deep dive on our video teleconferencing spend, video teleconferencing is very high end at the time, but heck we had it back in the 80s. When I was on active duty and it's still the same technology and they sit half the time unused. So you have a room with the cameras you're paying for but you're only using it for the weekly meeting. So we found a bunch of stuff like that and some of those are turning into savings opportunities in real money. We've been pulled off. Kevin this morning, Mark Krisco I believe it is. Talked about big G and little G when it comes to government, it's kind of the policy and leadership as it comes to this piece. How does that fit into the year? So the reason this has worked is we have a demand signal at the top of the organization. If the deputy secretary hadn't wanted this, we wouldn't be able to pull it off because I don't think we're different, but nobody likes to share their data, particularly when you know, in this case they knew we were going to use it to cut their head off. So we were going to use it to go cut off their left arm, have a nice day. So you need senior leadership involvement whether you want to call it governments or whatever in commitment to getting the data. And then you got to assemble a team that can actually massage the data, ingest it and then turn it into leaks. But then there's a little ROG of once you've got that all assembled in the data leak or what do you want to call it? How do you manage who gets to look at it and who can see how much? We have the trip, you know, Air Force can only see the Air Force stuff, Navy can only see the Navy stuff or do we want to go with a more open model with if you've got the time and the energy knock yourself out, babe. I've run a repository for years of all our IT applications and at the time it was novel but we said everybody can see everything. And then it turned out nothing bad happened because nobody has the time and energy to go chase everybody else's bad data, their own bad data is their own problem, right? So yeah, so that was that part of, that's the little G part that's pretty interesting though. We hear a lot about the skills crisis in big data, finding data scientists almost impossible for companies that have lots of money to spend. Government agencies don't have lots of money to spend, what are you doing about that? So we went and found a data scientist, right? And so we got our one, right? So I'm holding on to them real close and not letting them go. And I've been chasing for the last 18 months getting the mid tier one where they want to call it data engineer, whatever term you want to use, which can do our Python, do some parsing, do that kind of stuff. We have a whole boatload of business analysts that are basically Excel bound and maybe a Tableau or a click kind of level tool and working real hard. So I put, you know, we have the OSD workforce, the staff, or maybe not as old as me, but they're close. All right, so when you're looking for people who know our arms, Python and Scala, you're not going to find it in the government workforce. So we have to acquire it through our contract support. So you're putting pressure on your large integrating vendor to deliver people with a skill sense that have a clearance because we like our security stuff. So that then down checks the pool who are then willing to work given the tools that we give them. So, you know, Windows 10 and a five year old laptop is not real exciting for somebody who just graduated from MIT and has all the latest toys. So convincing them to come work for you or for somebody for you with a five year old laptop with four mega RAM and Windows 2010 is a bit of a stretch. So we've slowly been able to build out our bench here but it's taken a year and a half and it'll probably be the end of the calendar year. So that's really two years till I have kind of a minimalist set of skills across the spectrum from data science down to business analysts to kind of get this done. So it's just an agility issue, right? Can't hire into the government workforce, it takes forever. Then you got to go through the contacting route and then they have challenges too. And then they want to have their own data people to run their own business and then they have other customers. So prying the precious resource out and getting it on your contract as opposed to somebody else's contract is a challenge. So I think it's a global challenge. We're disadvantaged in terms of salaries and a few other things relative to what other people want to pay. We have this going on all the time. We have some of the world's greatest cybersecurity experts work in the Department of Defense in various places. They're making 165 at the top end of the pay scale generally, they can go to Wall Street and make 450. So there's got to be some other character, defense in the nation, some sort of incentivization to get people to stay around or the coolness of what they're doing. No magic bullets, I'm afraid we're out of time. Kevin Garrison, I want to thank you for joining us. I know we're all pulling for you. As I said, as taxpayers, we're all pulling for you to fund, make those six billion cuts happen. I'll do my best. And many more after that. This is theCUBE, we'll be right back with Steve Todd of EMC. I believe we'll be joined by George Gilbert of Wikibon as well. This is theCUBE.