 SiliconAngleTV and wikibond.org present Oracle Open World 2011. And now, host John Furrier and Dave Vellante on theCUBE. Okay, we're back at Oracle Open World. I'm John Furrier, the founder of siliconangle.com and we are here inside theCUBE, which is our flagship telecast where we go out to the most important tech events and talk to the smartest people we can find, startups, executives, thought leaders, analysts, anyone who's got some knowledge to share and we share that with you. We extract the signal from the noise. We are at Oracle Open World Live in San Francisco, California, where we are in a small little studio, kind of a gorilla operation set here with HD, pumping it out to you over justin.tv and siliconangle.tv. And I'm here with my co-host. I'm Dave Vellante of wikibond.org and we've got a great guest, Aaron Passi, CTO of Clustrix. Aaron, thanks very much for coming on theCUBE. We met last night with Paul, John, you were there and we had a good meeting with Paul, so thanks for being here. So Clustrix is a startup in San Francisco, California. They're a small company funded by some venture capitalists here in Silicon Valley in San Francisco and really are at the cutting edge of the database wars and the new generation of solutions that's powering the gaming business and enterprises all around the world and these guys are growing like crazy and the controversy really is around SQL, MySQL, NoSQL and you guys coined the term NewSQL which is a term you're trying to get out there but what's notable is you guys were the guys involved in founding Isilon Systems which was bought by EMC for how many billion days? Two billion. Two billion dollars. So this is not just some app guys doing some startups. You guys actually know what you're doing. You had a huge exit in your last company. You bring a lot of that DNA to this startup space where all the actions of the networking and database layer. So tell us what's going on with Clustrix right now and why are you guys so excited right now? What's going on with Clustrix? We are expanding our business. We're new customers, very exciting. A bunch of them in production. New features coming out. Scalable databases is continuing to... It's continuing to... You can spit it out. You can get it out. I can get it out. It's only 1,000 people watching right now. It's only 1,000 people watching right now. But you guys are in that whole SQL, NoSQL debate right now. You guys coined the term. What is NewSQL? NewSQL is the concept of scalable SQL. We don't want to have to require the application developers to throw away all their code that handled relational stuff in the past. We do atomic, we do acid compliance, we do atomic, we do transactions, we do relational calculus, all of that, and we do it in a scalable way. We can start with a small cluster that's only three nodes and then continue to expand as large as you want it to be, as fast as you want it to be. So there's a perception out there that SQL doesn't scale. Why is that perception there? And is it correct? Well, the perception is there because SQL hasn't scaled. You look at all the different systems that are, quote, scalable. I mean, you look at Oracle Rack. I have to say, Oracle would say, oh, we scale. I mean, help us squint through that, Aaron. I mean, Oracle Rack is a very mature product. It's reliable, it works well, but it has a limit to its scale. You look at, as you add nodes to it, you'll get an asymptote of how fast it can go. And when you add nodes beyond that, it doesn't get any faster. If you contrast that with Clustrix, we have a completely different architecture and when we add nodes, we get near linear scalability. We've done large-scale tests and have demonstrated that in the field and in the lab. Yeah, so there are, again, there's a perception that you got to go to, you know, NoSQL, MongoDB, for example, to get true scale. How do you compare to something like that? Can you scale as well? We actually scale a lot better. If Sergey Sarev is one of our founders, did a test a few months ago, where he took Mongo and he took Clustrix and tried to do sort of the same workload. It's a little bit tough because there's a lot more features with Clustrix and it's easier to make it work in an application world with us, but we had the same sort of app and we started to scale. And when you do reads that way, Mongo and Clustrix, they've looked about the same. They continue to scale. But once you start doing writes, you get this horrible lot contention in Mongo and it doesn't actually scale in the way that you would expect it to. So its primary purpose for being is scale and it doesn't demonstrate that in all the workloads that matter. Okay, so the obvious follow-up question there is, is Sergey a Mongo expert? I mean, he's obviously a Clustrix expert. Is he a Mongo expert? Is he able to, I mean, are there ways around that or is that fundamentally an architectural issue in your opinion? I think with Mongo, it's probably not architectural. I mean, I think it was the implementation. And I'm sure that that will expand, but you still don't get any of the additional features. They've thrown out so many of the features that everybody has come to expect in a database that it's kind of a different beast entirely. So, I mean, we're talking, we're getting a little bit down in the weeds here, but talk about the value of being able to scale SQL. I mean, it may seem obvious, you can maybe just drop it in and have to rip and replace your infrastructure, but talk about that a little bit. What's the business value? If you look at all the web properties out there, their dream is to have the entire world as their user base. They want to, they'll start off with, you know, a thousand. That's Oracle's dream too, by the way. Exactly. But they'll start out, they'll get a successful concept. They'll get it out there as quickly and as cheaply as possible. They'll use MySQL, they'll get it working, and then all of a sudden, they're the victims of their own success. So, they will start, they'll have various things fall apart. They'll, their users start getting a bad experience now that there's a ton of concurrency, a ton of different people on there. What clusters can do is it can go in and it can just replace that MySQL box just straight up with no modification to the application, and now you can scale it as large as you want it to go from then on out. And you can continue to add nodes to grow with the business. So as they make money, they will be able to give us money and put nodes into their clusters. And so if I'm a MySQL, actually we are a MySQL shop, and we haven't hit that scaling limit yet. Our dream is to do just that. But you're saying we could drop in clusterics non-disruptively, essentially. And be able to scale our business. Yeah, absolutely. You can actually put clusterics in as a MySQL slave using the MySQL replication protocol. It'll copy the data over to the clusterics box and then you can transparently flip it in as the master. And it's a very seamless integration strategy. So who are some of the people using clusterics? Can you share that with us? Who's using clusterics? We've announced several customers. Photobox in Europe, which is the largest photo sharing site in Europe. We've announced EBI as another one, which is actually a European Bioinformatics Institute. So we have life sciences type applications. We have the ladders, 100,000 K plus job site. I offer, which is like eBay here in the city. Right, okay. I'm waiting for Wikipedia to use it because we use the same software as Wikipedia. That's when we'll start using it. We actually have our internal wiki on our clusterics. You said earlier, victim of their own success. A lot of the enterprises have Oracle, for example. I've been living with Oracle for years. They trust Oracle because they're in there. Not so much trust them per se, but they're comfortable with Oracle, but they do pay a lot for it. You guys have a solution that has a scale opportunity for enterprise. What's your core value for people who might not want to go with Oracle for scaling out? Well, our core values are sort of threefold. They are scale, which we've talked about. Fall tolerance. In our system, any piece of the system can fail. Any node can fail. Any switch can fail. Any drive. And it transparently handles that without any users even noticing the problem. And that's a huge deal for many of our customers. And then the final thing is ease of use. We have an appliance model. You drop it in. It works out of the box. It takes 10 minutes to set up. It takes 20 seconds to add an additional node when you need more capacity. It'll automatically redistribute the data on the back end. There's no real trick to making it work. It just works. Every company has their own little secret sauce or some little thing that they do better than anyone else that they're proud of. Usually from the founders or the team. What do you guys do that clusterics that you guys say, we do this really, really well? That's different than you guys are proud of. And everyone, it's in the DNA of the company. Is it whether it's shipping code, managing hardware, fault tolerance. Is there one thing that you guys say, the company's DNA and the culture is we do something really well? What is that one thing that you do, you'd say? Can you share with us the insight there? Well, maybe just from the product point of view. We resolve queries in a fundamentally different way in our system. So back to the Oracle Rack. They have this shared storage back end. And what they do is they actually pull the data to where the query is running. In our world, we actually ship the query out to where the data lives. We decompose the query into query fragments. We send it out and make, and can in parallel execute that query plan. So you can actually, as you add nodes, individual queries get faster and concurrent lots of queries get faster. And this is really unique in the world basically as far as that goes. We're really proud of the way that we've taken this concept and taken it all the way to a fully functional database. So I have a question for you. So we are out the other night talking to someone from Riot Games and there's a lot of gamers out there, platforms that are growing like crazy, literally two million concurrence, 20 million Zynga has stats that says at any given time at the worst hour 60 million people concurrently on their servers. When does a company get to the point where they say, okay, I got to get clusterics in here from a scale standpoint. You said, you know, because most guys are putting up some cloud. When do you see clusterics really kicking in? And what are their challenges as they get to that level of scale? Well, before clusterics came around, companies like that would hit a cliff. They would, their one server would be dying. They would try to do some replication which can make some of the read-only traffic a little easier. But the only real way they had to scale before clusterics was to do starting or partitioning the data to multiple databases. And that was a huge, huge amount of effort from both a developer and operational point of view. So the ideal place to really consider clusterics is at that point, before you do all that work, because we make it so you don't ever have to and you can continue to scale. You can put your resources into making the game better rather than working with a backend database which doesn't really, your customers don't actually see that. The other thing about these gaming environments too is this happens also for applications is real time. I mean, people need to be, if I go to, we heard from Fusion IO CEOs is if Apple sells a game, they want to instantly have that purchase happen in real time. It's not a lot of downtime. So it's a big push to in-memory. What you got, how do you guys work with that whole in-memory trend and has it affect the clusterics by a proper, if any, does it? Well, the in-memory is, the in-memory databases are sort of a, it seems like a bit of a hack to be able to throw hardware at the problem. They, they end up with a- That's clear, here. That's a blog post right there. All right, so go ahead. I mean, there are applications where that's the only thing that's fast enough, but for the most part, we have the scalable system. We have a ton of cash actually. We have a lot of memory there, but it's all- It's like, not as currency, but like cash, cash. Memory cash. Memory cash. C-A-G. Kind of well-funded. We know that. You got some cash in the bank. Absolutely. There's not a ton of cash, per se. But everything that gets written to our system is persistent and atomic as it's written. When we respond that something is there, it's there persistently. And the in-memory systems don't have that property in general. You have a much more lazy way to get it out to something persistent or sometimes not at all. You'll just have multiple machines that if both of them die at the same time, then you've lost data. So in our world, we don't have these trade-offs. We can have the performance. We can have the persistence and make it a general-purpose database. Where's the idea come from? I mean, you guys, Icelon background, what did you learn from that? And how did you get to this notion of making SQL scale? I think it went out on a lot of customer calls at Icelon. And most of our customers at Icelon really loved the product and it really was an excellent product. They would go in and it would fix their unstructured data problem. They would go in and it would scale as large as they wanted. It would be as fast as they want. But we had dozens of them say, if you could do this with my database, I would have already purchased it. Because oftentimes the database for those customers were the bigger problem at their site. But they needed to solve one and they needed to solve the other. Icelon didn't have a solution to that. It was only scalable on the storage side. So after you hear that a few dozen times, it's like, well, maybe there's something there, right? That's great. So Aaron Passi, CTO of Clusterix. I mean, just fantastic story with Icelon, new startup, hot area. My last question is so as the CTO, where are you at right now with the product? Where are you spending your time? I mean, we're spending the time putting new features into the product, doing, you know, making it robust, finding new customers, doing research into different verticals to be able to find places where it'll fit. It's sometimes surprising where you find things to go. So that's, like I mentioned earlier, the BioIT, Bioinformatics, it looks to be a really big user of this. And we want to continue to expand where we're applicable. Excellent, well, thanks very much, Aaron, for coming up here on theCUBE and filling in for your CEO, Paul McKessel. We appreciate that, and thanks for coming on. All right, thanks very much. Appreciate it.