 Hi everybody this is Dave Vellante and we're here at the HackReduce with Adam Fuchs and he's the CTO of Squirrel, Adam. First of all welcome. Thank you. Thanks for coming on. Now I want to geek out a little bit with you, if you don't mind. Explain the tech behind the Accumulo project and what Squirrel brings. All right so what we're trying to do is bring big data security and analytics all together. Right so what we've done is we've taken Google's big-table design, we've extended it by adding cell-level security in and we've built a wide range of design patterns on top of that to build complex analytics right. When you bring all three of those things together you get a nexus right which allows people to just innovate right around big data right. Yeah so so the the the project came out of really a spawned out of big table right you guys saw a big table and said hey we can improve upon this is that right? Yeah right so we were working at the National Security Agency on this and we were trying to solve big data problems but also a lot of multi-level security problems so we looked at Google's architecture and said oh this is a pretty good start for our design right. We took the big-table paper that Google published and we started from scratch there built our own version of that threw in some techniques from a whole bunch of other infrastructure that we've been developing and voila you get Accumulo out of that. So you guys talk a lot about cell-level security and after studying this and talking to some people my takeaway and I wonder if you could confirm this or maybe of course correct it is really what you guys have is a a multi-tenant no-SQL database for the cloud at massive scale is that a right the right way to think about this? Yeah I mean that's part of it right but you got to think about what do people do with with big data in no SQL databases right. A lot of times you know people have a row-level security databases which will protect things that come from different sources they'll have column-level security right which protects different schemas but when people do big-data analysis they flip the data on in they pivot it right and the row boundaries the column boundaries they don't match anymore so cell-level security allows you to continue that security that that data-centric security model into the application space. What's the performance impact of all that you know security code? Talk about that a little. Yeah so the nice thing is that with cell-level security the combinatorics although you know they're possibly huge in practice they're actually pretty small so I might have a trillion entry database that's got a thousand different unique security labels right so we can use that fact to both compress well and also cache results of computation and it ends up being the cost of performance is negligible. So if you tried to apply this tech to a traditional database no SQL database it would it bring it to its knees? I think if you tried to you know got a couple it on top of it that would be a challenge yeah there's a lot of optimization that goes into a cumulative make it work well. Congratulations on getting squirrel off the ground we'll be watching I really appreciate you taking some time with us keep it right there we're right back