 Live from San Francisco, California. Extracting the signal from the noise. It's theCUBE, covering DockerCon 2015. Brought to you by SiliconANGLE Media. With special thanks to Docker. Now your hosts, Stu Miniman and Jeff Brick. Welcome back to SiliconANGLE TVs. Live continuous coverage of DockerCon 2015 here from San Francisco, California. I'm Stu Miniman with wikibon.com. One of the big things we're talking about at the show is really the ecosystem that's building around Docker, the project, the company, and happy to have on with me two members of that ecosystem. So first we've got Mark Davis, who's the CEO of ClusterHQ, the company that's the main driver behind Flocker. Thanks for joining us. Thanks, Stu. And we've also got Ken Dorazo, who's a VP in EMC's corporate office at the CTO. Ken, thanks for joining us. Thanks, Stu. All right, so both of you, first time on theCUBE, right? Yeah, absolutely. So Mark, I'm surprised you haven't been on in some of the various roles in the past, but thanks for those that don't know. We go out to all the big shows, help extract the signal from the noise, and thanks for joining us. Mark, you've been in these ecosystems and watched them grow before. Give us a quick snapshot of your background, what brought you to ClusterHQ into the Docker show. Well, so I've been doing technology things for way too many years, but I'll talk about my last company, which was a company called Versto Software, which was a virtual storage software play that was dealing with how do you optimize the way storage systems integrate with virtual machines? And had a long run as an independent company and company was acquired by VMware a couple of years ago, and that technology is now part of the VMware vSAN project. So when I heard about containers becoming big a couple of years ago, and we started to see what people were trying to do with them, we realized there's going to be lots of people trying to put containers into production environments, and when you think of production environments, you think of protecting your data in a very careful way. And what we didn't see when we surveyed the landscape was anybody really working on how do you make containers that have data in them reliable, fast, performant, easy to manage, et cetera. So that's what got us started at ClusterHQ, and that's what led to Flotter. All right, yeah, you know, remember, I mean, I think we all have, you know, the battle wounds from, you know, server virtualization was great, but boy, you spent a decade kind of fixing the things that broke in both the network side and the storage side. Let's pull you in. Tell us a little bit about, you know, your role inside EMC about your background. Yeah, so I work in the office of the CTO, and I lead a team of basically advanced research and development, and so we've been really at the tip of the spear in terms of looking at dockers, open source, and other types of things in this community, and contributing back in many cases as well, and we identified an area specifically within Docker that we thought and aligned fairly well with Flocker in our R&D perspective, and data persistence is kind of where we think that is going to be a big key role for EMC to play. Yeah, I guess when I think about the early applications most people are talking about have been stateless apps. Absolutely. So, maybe Mark, talk to us, what was the announcement you guys made today, bring us up to speed, and why partner with EMC? Well, a couple of things, so dealing, as you said, dealing with stateless apps has been what Docker has been all about, and you can do great things with containers as long as they don't have any state that needs to be retained, and that's all fine and dandy, but to capture the entire application, we're all building software in a microservices architecture these days, and what we hear is more and more people want to run the entire application, not just the stateless application tier inside of their containers, so what Flocker does is enable people to put state in their containers and then have the ability to move a container from server A to server B, and maybe move it from one cloud to another cloud, and have the data go with it, so this is the project my company's on, our mission is to make it where all the storage systems can connect into Docker, and by the way we connect in with Docker, then all those storage systems can be under the umbrella of an orchestration system, like Docker Swarm, Ormizos, or Kubernetes, and so when Ken and his team, when we met with these guys, I guess it was a couple months ago, Ken, we just saw a lot of opportunities to do some clever things together, so the announcement is an integration with EMC storage systems, particularly Scale.io and Extreme.io, which are spectacular systems, hardware and software, with the Flocker system, which is now part of the Docker plug-ins ecosystem that was just announced a couple of hours ago that we haven't been able to talk about until this morning, but having this plug-in mechanism where everything can become seamless is essential here, and that's why we partnered with EMC to bring in their storage system. Okay, and just to clarify, when you say we can move the storage there, is that like a V-motion or do I have to have a reboot, restart? Well it means, so today it's not full V-motion, so when we think of a VMware V-motion, we think of having a virtual machine that is running and picking that virtual machine up and moving it from server A to server B while it's running and with no interruption of service and any clients of that virtual machine don't even know that it moves. So we're not there yet with containers, we will get there, but we're not there today. What we are, where we are today, is simply being able to take a container, move it from server A to server B and when it moves, have its data go with it. So there is some restart time involved and that's certainly one of the things that we'll work on going forward is getting rid of that. Yeah, and this is kind of why we chose the two initial platforms for this, both the Scale I-O and the Extreme I-O, in that both of them are very nice scale out platforms and from an all flash array, looking at the very highly performant type of database applications that we see moving forward, the Extreme I-O was kind of the right first shot to put there and then the server side sand kind of SDS implementation for Scale I-O kind of, those are the first two that made sense from an initial offering. Yeah, server sand is actually our term that we use at Wikibon, so okay. We've done a lot of research on that piece. Newer technology's out of the EMC portfolio, so Ken, question I have for you, is are your customers asking for this today or is this kind of a research project that we understand this is where the puck is going and we want to stay there? What are you guys seeing? So I think as you kind of rightly annotated before, most of the work in containers has been around stateless, but increasingly we see more and more customers saying, hey what about these other types of apps that we have that will have some level of data persistence and what do we do here? And so this is kind of a, I guess you could say a forward looking or a preemptive move in this direction. All right, so Mark, my mention, this is Flocker 1.0 now, right? It is. So tell us what that means. We think it's safe for you to use it in production and if it breaks, we're on top of fixing it. So it is a robust system, we do have customers running it live, we actually EMC and cluster HQ have a joint customer that are now putting this into production. So we think it's ready. We've got lots more software to go. It is a 1.0 and we have work to do to go to 2.0 and 3.0, but it's a good first step and it is solid stuff that we think is production ready. Yeah, well, software 1.0 is pretty good. On the hardware side, I always said, make that first generation, hopefully throw it out or don't get it beyond beta and refresh it before you go, software is good. So we've talked about going from stateless to stateful. What about security? Because is my data secure? If I'm doing this, how does Flocker help or what is the security story? Well, so the security domain within containers is the container itself, right? So we're reliant on the security infrastructure that's built in the container itself. And Dr. made some statements today about things that they're working on in that realm. And so the tendency, so the idea of having extreme multi-tenancy is part of what containers are about. So for example, this customer Swisscom that was a joint EMC cluster HQ customer is building a database as a service where they have went to put thousands, maybe as many as 10,000 containers on a single server and do this for customers who expect that their data to be private and not have, this is database as a service. So the security model has to support that. And that's part of what containers are about is to have a box around the data itself. Yeah, and I think the Flocker team has actually put out some software that's in a really, really good direction down this path. And additionally, my team continues to look at new areas for data persistence, isolation, and security. And I think EMC is going to be doing a lot more in that realm as well too, yeah. All right, so I'm a little familiar with Swisscom from what they were doing in OpenStack. Yep. So they're a company that's used to break in a little bit of glass and transmentress thing. Anything else, what can you tell us about this customer? Well, so yeah, so what Swisscom are trying to do is deliver services to their customers throughout Europe. And they are very heavy and aggressive users of OpenStack. And in fact, this particular application is running on top of Cloud Foundry on OpenStack. And to deliver this service, containers are so obviously the right way to do it. Firing up a container that is very efficient and can be done in way less than a second versus having to start a virtual machine for a customer, it's just an obvious thing to do. And for the stateless parts of their application, it was a fairly easy thing to do. But these guys do know how to run things at scale in production. And any of us who have done this in the real world know that sooner or later, something is going to go wrong. Hardware is going to fail, I'm going to need to upgrade something. I'm going to need to move things around for various reasons. So for them to deliver database and a service that is A, scalable, which containers provides, but B, something that has some operational ability. So when that eventuality comes, that I'm going to have to move this container. They needed a way to solve that problem. Containers, Docker does not do that natively. It has no notion of containers moving around or having a name, a namespace that lives beyond the life of a container. So that's where Flocker came in. That's why they came and started working with us. And again, it's a first, but really important step to making sure we can do all the things in production that we want to do with containers, just like we do with virtual machines. So Ken, I think back to the early days of server virtualization, two big problems we had is number one, there was the IO Blender effect. So performance was really tough because I didn't have predictable ability as to where the workloads are. And secondly, things like snapshots and replications and all those pieces, it broke because I didn't have it. I had it at the one level, not at the VM level. So are we going to have these same problems with containers? What's EMC's position, where are we today? Where do we need to go? Yeah, I think that we're going to have similar types of issues. But the good news is that I think that the community is already starting to rally around where we're headed with more stateful implications with containers. And as we start to move more and more of those stateful apps, I think there'll be some natural developments in that area to work with that. Including, I mean, even right now, we're already having discussions with ClusterHQ on stuff that how do we make it better and how do we enhance both the data persistence story as well as tackle some of these really tough challenges. Yeah, this is somewhat like the server virtualization revolution, but it's different. Containers are not just merely a different way to virtualize. What's exciting about this is how fast it's happening. Right, I mean, you mentioned, it basically took us about a decade to get virtual machines to do all the things we want to. And we're still not done, right? There's still things, you know, that VMware is still just rolling out things. This is hard software and it takes a long time. What's really exciting about what's happening at containers is it's happening super fast. And I think a big part of that is because it is open source and there is no monolithic company controlling everything and I'm super excited about what Docker has announced today. We worked on heavily with them. This idea of having plugins. So now, you know, building something that allows third parties to plug in software so that it's natively manageable by the platform itself is a huge step forward. When we were at Verso software, you know, we had all kinds of cool things we could do to solve the aisle blender, right? We could make virtual machines go much faster. But because VMware at the time was a closed platform, you know, if you couldn't get inside of VM kernel, you couldn't add your software and only VMware could be inside of VM kernel, right? So it took longer in that environment and the fact that all this container stuff, you know, from LXC and Linux on up, is based on open source, is part of what's making this go faster. It's still going to take a couple of years for us to have it do everything we want it to do. But boy, it's going, it's part of the fun of what's happening in containers is it's so fast. Yeah, so Mark, you know, talking about open interfaces, obviously Flocker is working with EMC. What other solutions does it work with? So today we work with Amazon EBS as a back-end. So if you're using EBS as your back-end store, we also work with Cinder. So if you are in an open stack environment, anything that plugs into Cinder, okay, is manageable by Flocker. We also have our own implementation of a local file system. We've done a lot of work with ZFS or ZFS, as my friends in England like to call it, which we use as a local file system. We have some pretty clever ways of snapshotting and replicating using that. That part of our software is still experimental and is not released as a 1.0. But we have multiple back-ends we support now and we think we're covering a lot of ground with the ones we have. But part of what we wanted to do, just as Docker wanted to make it easy for third parties to write plug-ins, we have a pretty straightforward driver interface. So Ken and his team, it didn't take months or years to build these drivers. So maybe a matter of days slash weeks. And so for other companies or other open source projects to build a Flocker driver, it's not terribly hard. And there are others already in process and hopefully we'll have some others to announce pretty soon. We say one of the big differences we see between this wave and previous waves is it's happening so much faster. So the network effect of open source, so many companies getting involved. Actually, you guys did a survey recently, Mark. Maybe you can give us some of the highlights as that of what you're seeing from customer usage, adoption, surprises from that container service. Yeah, we were trying to figure out what's the state of container adoption in particularly in production operations, right? So, because we all know that with a half a billion downloads of Docker, there's something going on certainly amongst developers. We want to understand what's happening in really, really in production. So we contracted with devoffs.com to do a survey with their user base. And came up with some interesting data. First of all, surprisingly large amount of Docker in production. 40% of the people in this survey, and this is about a 300 person survey, so it's statistically significant, 40% almost are running something in production with Docker, which is pretty interesting. So that, and of course plans to go much higher in the next 12 months. The other thing that was really interesting to this conversation was what percentage of people thought that dealing with data was important. And again, as you said earlier, Stu, a year ago, nobody was talking about what we call stateful containers. People were assuming everything was stateless and state was something that you didn't deal with in your Docker environment. That was independent of Docker. And now, in this particular survey, 95% of the people that responded said that container data management had some importance to them as they thought about going into production. So what that means is what we're doing with ENC, and there's other people doing interesting storage things now with containers that are being announced here at this show. All this stuff is going to become more and more important, I think. All right, Ken, wondering, do you have any kind of early start points? If you're talking to customers and they say, how do I get started? What's your recommendation to them? That's a great question. That's a pretty broad question, actually. Do you mean with regards to just stateful containers or just in general? Yeah, so from a storage standpoint, I guess it would be the thing. There are certain apps that you're saying, okay, these ones are good ones to start with. These ones, you probably might not want your globally dispersed cluster of mission-critical stuff yet, you know? So there's a ton, actually. We're working with a lot of different customers on a wide spectrum of microservice architectures, as well as looking at ways to make databases like a Mongo and Cassandra, et cetera, a lot more extensible and more performant and deterministic, et cetera. Some of those things you mentioned before. I guess it's not, we normally don't sit back and guide customers into the right types of applications, but rather we take a look at what they're trying to do and then try to match them with the right types of technologies in order to manifest what they want, the outcome that they're looking for for their business. Okay, so we're running out of time. We're at the beginning of the show here. What are you hoping to get out of the show? Where are we with this whole way of life here and why is it so exciting? Mark, we'll start with you. Well, it is exciting to see the growth of this, right? So we were just talking about how big this show was a year ago, the Amsterdam show that happened six months ago and then this show, and a lot of people got turned away. So this show could have been bigger if they just had more room to see people. So the excitement is quite severe and what I'm really pleased to see is the number of engineers who are here looking for solutions to figure out how do I broaden what I'm doing? And Solomon Hikes, when he was doing his presentation this morning in his keynote, said the whole, he asked the audience, how many people here run Docker and only Docker? Not a single person rose their hand, raised their hand because of course it is part of a broader solution. So what I'm excited about is the number of people that are doing interesting things around containers. You mentioned security, there are some things to be done there. So I'm excited for people to work on that on those kinds of problems and I can't wait to go see all the booths of all the other people that are doing interesting things here. Yeah, and for me, I echo many of those. I think the ones that have kind of come out for me is really interesting are kind of the excitement around the community and the acknowledgement that there's still a lot of work to be done and there's a whole lot of avenues to come in and blanket and really help out the community to get stuff done. Yeah, Mark, as a closing note, I think when we talked last week about the announcement you were making, you were saying, all right, if we were looking at the VMware ecosystem, where are we? It might be at the virtualization, VMware desktop at this point. It's not ESX, it's not vSphere. We're back where we were a dozen years or so. All right, Mark and Ken, thank you so much for joining us. Have a great time at the rest of the show and thanks for sharing and being part of the ecosystem. This is Stu Miniman, I'll be right back with our next guest or right after this quick break. Thanks for watching us from DockerCon 2015.