 Okay, welcome back to the Cube, this is SiliconANGLE's flagship program, we've got the advanced instructor from the North Island, John Furrier at SiliconANGLE. I'm joined on this segment with David Floyer, co-founder of Wikibon.org, chief technology officer of Wikibon, and one of our favorite guests, I think he's made a Cube debut two years ago here at Oracle Open World, Brian Bukowski, founder of Aerospike. Welcome back. Right, thanks to beer. So in memory, what a great focus for Oracle. To be talking about in memory and big focus, what's your take on this? I was saying you're in the... What's my take on it? I mean, are you kidding? I've been a huge proponent of in memory from the day we founded Aerospike. So we saw it four or five years ago that rotational disk, the writing was on the wall that it was going to have a niche and that niche was going to be decreasing what we saw at the Flash Memory Summit a month ago from Samsung and Intel's roadmap as well is staggering and the writing's clearly on the wall now. Rotational is a smaller and smaller role to play and well-crafted software can really take advantage of in-memory computing, not just for key value store usage but also not just for fast analytics, but really for both. I wanted to bring that up in kind of tongue-in-cheek because one of the reasons why we wanted to bring you on the Cube early on was you were really early as one of the companies pioneering a lot of the in-memory when everyone else was kind of over the side. And so props to you guys, congratulations to Aerospike for doing that. But in all serious now, the debate we had earlier on with Dave on the Cube here was how early are we? Are we midstream or is it still early? Because Oracle's known to coming to the party a little bit just in time or late enough to be there. What's your take on that? I also want to get David Fleuys take on that as well. Well, let me ask a question slightly differently. Are they going to freeze the market for people like you with this grand vision of everything working together and everything being under Oracle? With the data stream processing and the analytic processing and the transactional processing and the big data processing, it's a big vision. Is that going to freeze the market for you or is that going to validate the market for you? So I believe it validates the market for us. We've been beating Oracle NoSQL in the field with customers for nearly a year and a half now. And that's based on our price, based on performance, based on not wanting to deal with Oracle's salespeople, based on the uptime and reliability we have. Reliability is huge. Once you have customers like we do running thousands of SSDs and large-scale deployments with zero downtime and tens and hundreds of terabytes of storage, that's the kind of lead that we have in terms of actually being out there in the marketplace. So Oracle is also, as you're saying, perhaps just about in time. They've caught along on with this at a point where folks are really starting to say, hey, what about that NoSQL stuff? I really got to learn about the power that in-memory flash computing can bring and start to actually use it. So I think they're hitting the market at a decent point, which their product was a little bit better, but they'll get there eventually, right? We don't know much about it yet, do we? Well, it's been out there. Yeah. Right, so Oracle NoSQL's been out there. They finally have their JSON interface and we'll see how it does. But what about other approaches? If you look at Big Data this year, it's amazing the number of SQL people that have come to the table and people like Clustrix who've got a very nice story with a flash, similar flash architecture to yourself who are going down the SQL path. Their reasoning being is that we can make it almost as good as NoSQL and that the programming is simpler and it's just easier to implement and it's easier to find the talent to do that. And they'd rather, especially in the enterprise space, not leave tidying up NoSQL out-of-line situations to the programmer. They'd like the database to do that. But any comments on that is do you see that as a threat or do you see that as a broadening of the marketplace? So the experience I've had when talking to enterprise customers who have a lot of installed-based SQL is that at some point they have to bite the bullet about whether they're going to stick with that SQL environment, SQL programmers or jump to an environment where they've got some Hadoop and they've got some NoSQL, right? And what I see is that at some point it's really, really hard to make that SQL scale for a bunch of reasons. And Oracle is finally getting past some of the simple implementation problems that they've had with NoSQL with their SQL implementation, right? And getting on the board with Flash and their Flash analytics, but they've still got the problem where you can express things in SQL, SQL is extraordinarily powerful, such as these large scale joins that simply will never scale with that kind of expression where it's very simple and everyone knows now how to build a boot cluster and how to build a NoSQL system that works properly. So David, I wanna get your take on Oracle's move. What's your assessment of the in-memory take? So my assessment of the whole Oracle strategy is that they want to carve out a single platform, both from a hardware point of view and a software point of view and integrate everything as much as they can. There's clearly a lot of SQL customers that they have. There's clearly a requirement in some organizations not for the level of data streaming requirement that many of the web-based people have, but for web streaming of some sort, streaming of some sort, not quite the same level. They'd like to integrate that in with analytics, do that in near real time, get that moving into it. There's clearly a demand now for that end-to-end type of system going in. And if they can make it work and if they can make it attractive and if they can package it up in a way that minimizes the cost of maintenance, et cetera, they will take a chunk of that market, but it's all gonna be about execution over the next two or three years. They can probably hold people off for two years, but beyond that, people will have to move. So I think the next three years is gonna be a very interesting time in the Oracle base. So I already see that there are plenty of folks who are out there building large adobe clusters and figuring out what's put in front of it. I think some of the most innovative work I'm seeing right now outside of the advertising industry is within both security and that there's a lot of different kinds of security out there, right? So there's transactional security for banking. There's fraud in advertising. There's fraud in online gaming. There's really fraud everywhere you look. And once there's money, there's fraud, right? So the ability to put together a adobe cluster that is looking for the big picture patterns. If I see this, that probably means fraud and then have a real-time flash-oriented no-SQL database really on the front side looking at every single moment of what's going on. That's what I see really winning the fraud battle. And SQL was what people used in fraud. Still I would say 80% of the market's using it, but those companies that are out there and have placed a bet on Hadoop are doing so much better job in detecting fraud at a reasonable price. It's hard, I think, for an architect right now to say, I'm starting a new project, I don't have a SQL base, I'm going to build this whole thing on SQL. People don't do that. They just start with Hadoop these days and then they put a fast database in front of it. That is the architecture right now if you do not have legacy code. I think that's very fair, but there are also people that are going to come the other way. Not starting with their problems, starting with their traditional system and wanting to add on the fraud detection and other things to that in a way. And the challenge of some industries, you can lose a piece of data and it doesn't matter. If you lose a bit of web data, it's not the end of the world. If you lose... Well, if I'm playing a game and I just spent two hours getting my little pony in Zynga and I lost my pony, don't tell me I'm not important. Dude, my pony. Come on. Just the web? If I lose a data and... I lose a penny in my bank account, that doesn't matter either. Well, if you don't reconcile properly in your bank, then they're a big difference. Look, the other point that people forget is banks don't really use SQL for their mainline transaction processing either. SQL isn't considered to be reliable enough. They use things like XA transactions, they have multiple transaction workflows and their databases make mistakes and then they go back and reconcile them. So even in those systems, they really look a lot more like IMS and they look a lot more like what we now call NoSQL than they do a traditional SQL story. That's the highest level of reliability is what one analyst at JPMorgan called forming pre-SQL. We use pre-SQL and that's what's reliable. So what's up with... Give us the update on Aerospike, what's happening? Guys, you have a new CEO, what else is new? So things are going very, very well, John. So at Aerospike, we recently released Aerospike 3 so that has secondary index capability, real-time map reduce, so being able to map reduce in millisecond range times over decent sets of data, use combining both the ability to index and the ability to run map reduce over it. So not just getting your fast user data storage but also the ability to do analytics on top of it with all of our usual capabilities in terms of flash support and flash integration with people like Violin and Fusion IO who I guess were guests here earlier as well as Intel, Samsung, Micron, all the usual suspects. What else is around the horizon for this industry that you see that you guys are working on that you think are, as a start of yours at the front lines? What do you see around the corner? So right now, concurrent with this is the high performance transaction processing symposium and those guys, there's a very interesting talk or they're earlier today about the future of data centers and data center integration and the entire idea that we can use much faster interconnect technology to build faster and better distributed systems, whether it be PCIe, there's different forms of new fast ethernet going way beyond InfiniBand, that's gonna really change the nature of distributed systems and allow us to build things we've never built before. When you start looking at four to five microsecond response times between systems and whole data center distributed systems as opposed to these computer things, it's a whole new world. So that'll be rolling out, that's pretty far looking, that'll happen probably over the next three or four years. Honestly, the entire industry has to adjust to and digest what's happening here at Oracle Open World and being able to bring together NoSQL, figure out how MySQL and NoSQL play together, how Hadoop works together. That's gonna take us a few years. David Florial, let's hear the final word in software-led infrastructure. Io, send your give-a-share as you're going back to your early work. What are you working on right now? What are you watching in this space on the infrastructure side? Well, the thing that I'm watching is very much what Brian was talking about, the potential of putting together systems of such size and potential amount of data that they cover, that they offer ability to design systems in a completely different way. Absolutely design systems with huge amounts of data, take out cost from the infrastructure of organizations, make single copies of data a reality in the marketplace, making them both a databases of record as well as databases of analytics as well. I think that the amount of change that's going on in the industry at the moment, when you consider that memory is changing, when you consider that the internet interconnect is changing, when you consider that the amount of processing power is changing, we've never seen such an enormous influx of change and capability and innovation that's happening in the industry at the moment. It's as if the dead weights of the disk technology which was pretty well holding everything back are finally falling off. The dead weight of disk technology that's been holding everything back as David Floyd says in memory, flash persistent, changing the software games all at the software. This is theCUBE live coverage from San Francisco. We'll be right back after this short break. Stay with us. Day one of three days of wall-to-wall live coverage Oracle Open World Live in San Francisco. We'll be right back. I'm John Furrier. We'll be right back.