 Live, from the Julia Morgan ballroom in San Francisco, extracting the signal from the noise, it's the Cube, covering Structure 2015. Now your host, George Gilbert. We're back, we're live at Structure 2015 in the Julia Morgan ballroom in downtown San Francisco. We have with us another very distinguished guest, Derek Harris has been involved in the content and organization of pretty much all the Structure conferences going back eight years, pretty much. That's say 2009. Okay, seven years. All right, I was never a math major. That's all right. I wasn't with Gigo, I'm with for the first Structure conference, I was just an attendee. Well, so Derek tell us, so you've got a pretty long, a longitudinal view of how these conferences evolved, maybe tell us, starting back in 2009, what the conversation was about then and how it's evolved since. Yeah, so I think the conversation has been, and I'm sure other guests might have referenced this, but it was very much this discussion early on of, Werner Vogels would come for the CTO of Amazon and do this hard sell on Amazon web services and cloud computing and there's the notion that this is why you would want to do, this is why you'd want to run stuff in a cloud and this is why you would want to outsource all these things and this idea that, it was just a new concept, the idea that you would want to do this and then people talk about security and this analogy like well you keep your money in a bank so naturally you would want them and you trust the bank so you should trust this cloud provider and there was this very, just this early discussion of, is this a thing, can we trust it? But what was the alternative, or what were the alternatives at that time? Well they were VMware, and they were no sort of virtualization layer at all. Hug the server. Yeah, and so I mean those were the options and it's funny because yeah and today, the discussion is not like why you would do that is, what's the best practice, it's like you're going to do that, what's the best practice and how efficient can you do it and you know I mean, or it's maybe just how much, but it's not, yeah it's not even a question anymore that you would do that. And then I think another thing that I thought was fascinating actually was, we had a lot of talk about databases and no SQL databases and stuff at the early structure conferences and there was this debate, well SQL can't scale and no SQL, here comes Couchbase and MongoDB and Cassandra and all these things are coming around and they're going to oracle so dead. But then in time, I mean this is kind of the evolution right, but then we've come back and I think this is the case in a lot of things. Some of that excitement I think was necessary to get attention, but then it was dialed back, right? Or co-opted to some extent. To some extent co-opted and to some extent you said, okay yes, these are legitimate, it's not as if this is a stupid idea but we're still going to run my SQL. I talk to so many companies now it seems like or you know I would just say so many, but enough and some everyday type name companies and they're like yeah we tried this, this, this. We're running on like Ruby and MongoDB or MySQL now because it works and people know it. So all things kind of come around I think and so we're very excited and I think we dial it back as kind of a thing that's happening. I want to touch on that mostly because I mean databases are central to all applications and they sort of almost dictate the architecture of a core part. And the NoSQL databases basically rose up for two reasons as I understood them. One was they were more scalable cause they relaxed the constraints of rigid transactions. But they were also more flexible because they didn't make you spell out exactly where you're going to put your data before you put it in. But people moved away from MySQL because MySQL never dealt with those two realities. What are, how are people using MySQL and still accommodating that scalability and flexibility? Well I think what you've seen is over the years I mean the flexibility part is probably, I mean it's still an issue to some degree. I would imagine but you know if you have if you use cases if you know what you're doing enough I mean and MySQL makes sense it makes sense you're going to do it right now. The scalability question I mean there's been a lot of you know this was termed new SQL that was kind of quoted companies trying to do this but I think we just look at a Facebook I mean any large web company now is pretty much still running MySQL as far as I can tell. And they just put their own layer on it. They're doing all sorts of things to make it scalable. I mean Facebook had a whole team around you know my scaling MySQL. There's a project that I think was announced last year it was called WebScale SQL and it was Facebook, Twitter, Box, Google all these companies said hey here's what we learned and we're going to do this project. It was just funny to me all the excitement that happened Facebook invented Cassandra right? But then you know kind of you know the excitement wears off after a point and you get back down to what's real and I don't think that's the only technology you're going to see the same thing happen with. So let's go to other technologies. So we saw the sort of scalability and partial co-opting you know of SQL or of the NoSQL databases. What were some of the other things that came along that perhaps looked like they were going to be if not a panacea you know a big fix but that didn't pan out? I mean I think you could go back and talk about then this is still evolving but the hybrid cloud the private cloud open stack was a thing when open stack came along people said oh my God this is you know look at these this is going to we need this alternative to Amazon and VMware and this is going to be it and look at all these companies backing it and then in time well I mean I don't even think we need to go into doing much detail it hasn't proven to be that panacea in the most cases. I think containers are going to be a thing where there's a lot of excitement about Docker and Kubernetes in these things. Now and understandably so on the other hand that's not you know the whole world I don't foresee moving to Docker containers anytime soon. You know when we talk about Google runs Google still runs you know it's resource isolation it's containerized but it's not Docker right it's not as if that notion of containers begins and ends with that sort of thing. So I think we need to have just a broader view and take these things and really understand you know I think it's people need to think about these things they need to and this is why I like these early discussions but you know then when it comes down to actually putting it in production and building companies around it I think I think you know you pull that's when you start seeing be pulled back a little bit. Okay so looking out a couple of years what are the things and we know you're partial to mesos but what do you think the world would look like in a couple of years when we come back and you know talk to you including you know how many mesos might remake things. Well so I mean I think also I mean pardon my bias but I'm sure you've had other people from companies that I'm talking so I'll give my spiel. I think that if you look at what mesos and mesosphere are trying to do I mean you know the idea is essentially to build what we call a data center operating system but the idea is you take all your resources you look at them as one big computer like Google does by and large and you start programming against Intel's APIs and so you stop and you don't have the notion of well I had a VM so if I had like whatever four if I had four servers and I apply take I use a hypervisor on that now maybe I have 16 servers right and so I have 16 servers to manage our idea is I have four servers now I have one computer it just has the resources of all four servers and I'm going to write a job that says not I need this many nodes or machines I'm going to write a job that says I need this much CPU this much RAM this much storage and you just have to software deal with where that goes you don't have to spend your time dealing about that and so beyond that then it's this idea well one thing we'll do so you can install you want to run these just complex distributed systems yeah that's it's just easy to install it's much easier we'll run Spark here you can run Cassandra here you can run Kubernetes here you can you can use Docker you can do all these things you want to do but the the the interaction for developers and engineers and everyone it's just it's a much simplified flatter sort of existence so all right we've had a Derek Harris with us another distinguished alum of the of the old Gigam and and the resurrected structure conferences and we hope to see him again soon at future coverage at structure data and structure data and then structure connect we will be right back in a moment this is George Gilbert and we'll see you in a few minutes