 Okay, we're back here live in New York City. This is Big Data Week Strati Plus Hadoop World. This is theCUBE, our flagship telecast from SiliconANGLE.com. We go out to the events and extract the sealant from the noise. I'm John Furrier, the founder of SiliconANGLE.com. Enjoy with my co-host. I'm Dave Vellante of Wikibon.org. Go to Wikibon, check out the free research with IT practitioners like yourself and collaborate with them. We're here with Camille Fornia, who is the director of engineering at Rent the Runway and also an Apache Committer. Welcome, Camille. Hi, thanks for having me. Thanks for taking the time. Coming on theCUBE, Rent the Runway. So, pretty interesting, I'm on the website now. Check out Rent the Runway. Tell us about the site and the company. Sure, so Rent the Runway, our business is basically short-term rentals of designer dresses and accessories. So you go to our site, you have an event coming up, you go to our site, you search for things available on your date, you order something, you come to your door, either the day before or the day of your event, you wear it, you get lots of pictures taken, you have a great time, you put it in the mail and you ship it back to us. Yeah, so all of you folks out there that would go buy a dress, wear it, and then say, oh, take it back, you know. Everyone's seen it already, everyone's seen it already, can't wear it twice, right? That's right. Is that how you came up with this idea? Well, you know, our founders are two women, Harvard MBAs, and yes, I mean basically, really one of the sisters of one of the founders was like, you know, I really want to wear these gorgeous designer dresses, but they're so expensive to buy, and I wear a dress once, I buy a nice dress, I go to a wedding or I go to a party or whatever, and then my pictures on Facebook and every single one of my friends has seen me in this dress, and so can I really wear it again? You know, I don't really want to wear it again, so I'm just like buying these dresses and then wasting them and not wearing them more than once, so why not spend less money and rent something way nicer than I could afford to buy in the store? And look great. Yeah, basically. Social media, changing the disrupting the apparel business on the high end. Absolutely. So take us through some of the workflow, and I'll see if people are interested in this product so that you guys have a back end, you're going to build this house. Take us through some of the geek tech behind the scenes. Yeah, sure, so I mean, it sounds, I think at first blush, like a sort of simple e-commerce kind of problem, right, you know, you build a website, you put some merchandise on it, but we actually have a number of challenges. We're a very data-driven business, so first of all, you know, when we ship items out, we're not just shipping them out and then they're sort of out and they're done. We're shipping them out and then they have to come back. Every single item in our warehouse that ships out pretty much is coming back and it's very likely that when it comes back, it's actually needed very soon thereafter for another customer's order. So there's a huge amount of technology that we actually have to build to run our own warehouse. So you cannot buy technology off the shelf to run a warehouse where you're shipping out and returning high-value items like this. So, you know, one of the cool geek things behind this is really like just the software that runs our fulfillment system and that, you know, tracks all these items as they go through the stages of the warehouse, as they're cleaned, as they're repaired, as they're shipped out to other customers. And what does that look like in Hadoop? I mean, is it batch? Is it real-time? Right, so that's a very good question. It's a lot of it is fairly real-time and we're track scanning all these items real-time throughout their sort of lifecycle of events in the warehouse as well as outside of the warehouse. But, you know, right, it's not quite Hadoop scale, right? You know, this is big data but not Hadoop level. Not at a bite scale. We don't really need to do a lot of map reducing. But one big challenge we do have that we do in a batch way is every morning we have to sort of decide from the inventory we have in the warehouse, where's it gonna go, what's needed, what isn't needed, so we can optimize the human processes around actually getting that inventory shipped and we can optimize the human processes around preparing the inventory that's returning that may need to actually go back out that same day to get it out, get it cleaned, get it ready, get it going. So one of the things we've been talking about yesterday and today all day is obviously analytics, insights, and also business value, right? You can talk about data-driven business. Awesome topic. Let's drill down on that. So what does that look like for you guys? I mean, you have to have that kind of analytics. Is it like looking at how customers use the product? How they use the dress? Is it damn? All this stuff. What is your biggest data problem that you have to get your arms around? I mean, so I think we have so many data problems. I think one very interesting one is actually the way that we choose our inventory is very data-driven and it's data-driven in a way that's very different from a traditional retailer. So a traditional retailer, they will usually have maybe a budget to spend on particular designers. So a designer that's doing really well, they have a certain amount of dollars to spend to purchase into that designer's line. We actually do attribute-driven buys. So we look at, because we have this clickstream real-time information about if we had infinite inventory, what our customers would really like to have based on what they're viewing on the site, what they're clicking on, what they're liking, all of that kind of things. And so we can actually determine, based on their demographics and their clickstream behaviors, what kind of attributes, and that's things like dress length, so gowns versus party dresses, color, things like sequins, sleeves, what kind of attributes are popular now, are popular with our customers, what we predict will be popular, and actually buy into inventory based on inventory that matches those key attributes. So we do a lot of data analytics around inventory buy and attribute-matching to really optimize what people are going to want. Can you talk more about the predictive analytics piece of it? How do you do that? How do you go about it? Is that secret sauce we can't say in the game? It's a bit of secret sauce. So give us the high level then without divulging. So we have, you know, it's really just a combination of, honestly I can't talk too much about it because I'm not an analytics, I'm an infrastructure engineer, more than an analyst, I'm a director of engineering. Our analytics director could probably talk a lot more about it, but you know, I know that the various aspects that go into it are really like, we are looking at the clickstream very frequently, but it's a very heuristic based approach, right? So one big thing that we want to do moving forward with data is be able to, you know, instead of just sort of having these heuristic based data algorithms and using those to, you know, drive our business to actually do a lot more real modeling of, you know, the algorithms and real tuning of those algorithms based on the data in a much faster, tighter loop. What about databases? Because obviously databases are big. What are you looking for? What do you use for databases and how's that layout? Yeah, so right now we have a few different sources of data. Obviously MySQL is a big one. We have a lot of transaction data. That's very important. And you know, our analysts, our analytics team really uses a lot of MySQL. We actually use a lot of MongoDB as well. We've started doing a lot of that because we have a lot of sort of document structured, slightly more unstructured data, especially around our dresses and that kind of thing. And then we actually just have a lot of, you know, sort of straight log file data that we need to write, you know, code to parse. And that's things, again, like clickstream, log files, all that. Great, so my question for you, kind of to wrap up for the segment is your vision. Let's drill down and spend a few minutes on, say your vision, well, your vision, but you see all this stuff evolving from all this new stuff. We had platform on hot new startups as they do BI without ETL and data warehouse, which we love that. We love that. It's very disruptive. So what's that disruptive out there that you like that's a vision to where the data driven businesses are going? Because again, you're a great example of, it's not huge petabyte scale data, but it's actually real data. You need to get your hands on, play with, integrate, mash up. What's your vision of some of the new things that are disrupting? I mean, I think, you know, for me personally, the things that I'm most interested in in the space, you know, that I think we would make a lot of use of, is really a lot of the like, service vendors. I don't want to have to run a Hadoop cluster myself to get the value of this, you know, big data analytics platforms, you know. I just, I want to be able to sort of, I have all this data. I want something that's going to help make it easy for me to combine it all together, to put it all together, to write various algorithms on top of it, and to, you know, serve that data back, without having to hire a huge engineering team to support a hardware installation. I think that's, you know, I think from what we're looking for perspective, that's just so important, you know, whether it's disruptive, I don't know, but it's, you know, it's certainly something that we find to be extremely valuable. Can you talk a little bit more about the changes that are going on in infrastructure from your infrastructure engineering perspective? We've been working on, you know, the future of infrastructure, and how it supports this new data model, and can you talk about, and we've talked about software-led infrastructure, can you talk about what you see as the big change in infrastructure today, and coming in the future? That's an interesting question. I mean, I think that, you know, we're still very much in the babyhood of the NoSQL space, and the Hadoop space in general. You know, the companies that are really successful that they are either very big, and the Googles, and the Amazons, and the LinkedIn's of the world, or they just absolutely sort of have to be to do their business at all, sort of like Foursquare, for example, right? Specialized. You know, they're very specialized, and I think that, you know, every year you watch as the Hadoop infrastructure becomes more, it just becomes better. It becomes more polished. We have better tooling around the infrastructure. We have better services, better vendors supporting it. We have, you know, places where you don't even have to install it yourself at all. You could just, again, use it as a service. I think the as-a-service model is expensive, but it's so valuable for smaller companies, and I think that's just incredibly, incredibly important. And so I really think that, you know, I really think that the most valuable resource that a small startup has is human capital. It is the people that we have working for us. If we have to spend a lot of human resources on sysadmining, Hadoop clusters. Setting up, managing. Setting up, managing. It's just, you know, it's just dollars wasted or human capital wasted that we could be spending really writing cool algorithms like helping women shop by women like them, or, you know, understanding our data better. And so that to me is where the real innovation and value and growth of the community really needs to be. I think that's a really great point. The people-centric view, the human-centric element is very key. The simplicity around the tooling is absolutely going to be there. A lot of work to get done, for sure. My final question, final, final question is for the folks watching who aren't here. We're live at Strada. It's a packed house. They're kicking people out of the door. Not enough room, fire hazard, all that stuff. Explain to them what's going on here. What's the vibe here? What's the community like? Give them the geek view of like, hey, this is what's really going on here. Sure. I mean, I think what you're seeing this year, so I was at Hadoop World once. It was just Hadoop World last year. And the difference between last year and this year is really like the full adoption and acceptance that this is what everyone has to do to some degree or another. And so you see that in the attendees, where you see everyone, you see big businesses, you see small businesses like mine. You see just a huge variety of attendees. You see a huge variety of vendors offering lots of different services around big data. I think the vibe that you really see here is we all have really gotten a clue. Data is important. Data is driving our businesses. Data is helping us all make decisions. And now we're all really grasping for, okay, but this is still really hard. How do we make it easier? How do we think about this? We're all so comfortable thinking about old things. We're comfortable thinking about MySQL and LAMPstack systems. But none of us are quite at that comfort level with Hadoop and the big data processes yet. So how do we all, from enterprise to startup, really get that comfort level? As a follow-up to that, I want to ask them that's more of a small, medium-sized enterprise, kind of tech enterprise, or a small, medium-sized enterprise infrastructure question. In the old days, you get a knock on the door, hi, I work for HP, I work for Dell, I want to sell you hardware. Yeah. Right, so do you look at that anymore and saying, I don't really want to talk hardware. And then you got to knock on the door. I sell software, package software, SAP sells to the high, high end. But I mean, when you get those knock on the doors, are you saying, look, people, I want solutions? So how do you, as a small, medium-sized, because you're like the perfect example of enterprises that are small, medium-sized and growing who have, make decisions about, like, hey, my waste of human capital, I'm not going to do it. That's one example. But relatively, the guys knock on the door selling you stuff, which are mindset besides getaway. I mean, I talked a couple of minutes ago, we're a hardware company. I'm like, well, I don't think that's the world anymore. So what's your share with the folks out there watching, you know, the HPs, the Dells, the IBMs, everyone else? No, I mean, I think, again, I would love someday to be racking our own hardware. I want to be the size where it makes sense for us to rack our own hardware and hire the people to do that. And, you know, because I do think that you lose, you know, it's expensive to run cloud versus, you know, your own, maintain physical bare metal. But, you know, for the foreseeable future, I'm going to be in cloud and I'm interested in people that are offering me full-stack services. I'm interested in, I, you know, I have some need, we use Heroku for some things, right? I'm interested in basically being able to say, you know what, I want to be able to easily deploy my HP patient, easily scale it up as I need it, scale it down when I don't, have that really, the nice tooling that makes it so that, you know, my sys admins barely have to think about anything at all. I'm interested in Hadoop vendors that are like, as long as I can get my data to you, you've provided all the tooling around it and I just, you know, use an interface, a nice interface that you've helped provide where I write pig queries and I get information out of that data and I'm able to just take any of my analytics team and say, you know what, take a week, learn this product and go and not, you know, I'm just not interested in having to buy every single piece of the solution separately, put it all together myself and maintain it. Yeah, and then at some point when you, like you said, when you hit that milestone of, hey, you know what, it's economically to stand up some servers and roll your own data centers and or bare metal. Yeah, absolutely. All right, Camille, thank you so much. I mean, this is exciting. I know we've got the planes are backing up at the airport as we say here. Thanks for coming on. Great to have you on theCUBE. Your great insight. Love the small means of enterprise. I love this data-driven business concept and you're a geek, you're in there spinning up cloud the way it should be. Congratulations. We need all of your respect, by the way. Don't worry, I call myself a geek, so. No, no, it's a call man. Dork and Nuri, that's a whole other story. I haven't called that too, but that's okay. Alpha geek. Alpha geek, you're an alpha geek. It's just theCUBE, we have fun and we have a lot of signal here from you. Thank you very much. That's what we do. We start to see them from the noise on angle.com. We'll be right back with our next guest after this short break.