The Exchanges

Q Terrific. So, what are the fundamental advantages of a NoSQL and NoSQL database of our relationals? You mentioned scaling horizontally. What, what, what does that mean?

A Yeah, it, it, It basically just means that you can take a large number of servers and make it one database cluster. It looks like one logical database. It looks like it's one machine, and they work together. So we can take a hundred servers, say that's my database ABC, and, but I have the capacity in the, ah, both in storage and computation and memory and RAM of, of a hundred servers, not of one. And with traditional databases, it, it's, it's hard to get them to scale horizontally, as we like to say. Um, you know, the traditional way to scale, if we go back, say, 20 years, is vertically, right? You could buy a bigger machine, right? You know, you have a mini-computer, you can buy a mainframe. You know, you can buy a SunBox. So, you know, at DoubleClick, you know, we, ah, there are a lot of pieces of the system, but we used Oracle for certain things, like for storing, ah, business objects, like orders and customers and things. So, um, we just bought the biggest boxes Sun made, and had a couple of those, and then, but then we would, Kind of run into that ceiling then, and, but that ceiling, it's gotten worse because computers are faster now, of course, but the delta between sort of a random kind of commodity, um, Server that would be in a data center in a rack, and the fastest box you can get that's a single server, it's, it's, it's a lot smaller than it used to be. And just beca…

AI assessment note: “It basically just means that you can take a large number of servers and make it one database cluster.”

Q So a couple more questions and then I'll open it up to two people. Um, um, If you were going to start something today, um, let's say, you know, in the big data world, what, what are some of the areas that you think are interesting versus, you know, played out for people starting about, thinking about starting their own company?

A So, um, there was, uh, I actually saw this on Business Insider a week or two ago. There was, ah, I forget who it was, someone at Sequoia had said that, ah, um, entrepreneurs right now are ignoring B to B. Jim Goetz. Jim, yeah, ok. So, I read that and I thought, he's right. Um, so I think there is an underweighting of activity on B to B because there's plenty of opportunity there. And, you know, businesses need things. They, they spend lots of money. Um, they have problems to be solved. So, My first thought is look at B to B, right? So, cause it's a little bit of like a game theory thing. It's sort of, if you're trying to solve the same problem as everybody else, and there can only be one winner or three winners, then it's going to be hard. If no one else is working on the product you're working on, and it actually is a real space in the end, you know, your probability success would be higher. So, um, I think it's partially that. You don't want to Start working on a new idea, and 500 people are simultaneously trying to do that new idea at the same time. And, and I think the, the B to B space is, is sort of the underweighted case right now, competitively kind of at the startup, um, phase.

AI assessment note: “My first thought is look at B to B, right?”

Q So there's still a number of cases where, for example, if your data is relational in nature, um, having a relational database still makes sense? I mean, NoSQL, NoSQL is not positioning as a total replacement to relational database, or is that not correct?

A Yeah, so, uh, not a total replacement. I think one size fits all is over. But, um, my belief, I'm a little biased, but my belief is that if we kind of project out Say, um, five years, 10 years, is that what we're gonna see is you'll have a few tools in your toolbox for databases. Not 12, but more than one. Three, for example. And you'll have a few niche things too, right, for special cases, but standard things you use all, all the time. I feel like these NoSQL products could, it, it, it's gonna turn out that they are applicable to at least as many use cases as anything else you're using. So not everything, but they're not a niche. Their general purpose, at least some of them are, and I think that's our goal with Mongo. And so it will apply to as broad a swath of things as anything else does. And you're right, so you have to look at your problem and, and, and, and what tool is the best fit for it. So, And the data model would be a good one to look at. So if your data looks like documents, if you're, you know, if you're building a content management system, it would be sort of a no-brainer to consider something like Mongo with a document or a data model. Or if you have semi-structured or complex structured data, that can work well, too. So, for example, if, um, let's say we want to set, um, store, um, derivative securities in a database, history of traits. How do you represent th…

AI assessment note: “Yeah, so, uh, not a total replacement. I think one size fits all is over.”

Q Could part of the problem be that, um, it's harder to raise money, um, as a, at the seed level? I mean, you are in a completely different category because of your track record, right? But, uh, if you are a, you know, first time entrepreneur trying to do enterprise taking New York, uh, is a hurdle, the barrier?

A I think it's changed. I think it's not harder now. Um, so, which is good. I think New York has been underserved, at least for this, for our, for the IT tech type world. Maybe not for, um, healthcare or something, venture, but, um, in, underserved in terms of sort of the number of, kind of, venture firms and VCs in New York. Right? It was underweighted, historically. But I think that's changed. Um, I think everyone's paying attention to New York. All the West Coast firms are. Boston firms. See people moving here, like from Boston, right, who are VCs and they live here now, and quite a bit of that. So, I, I think it's totally fine now. Um, and

AI assessment note: “I think it's changed. I think it's not harder now.”

Q the product, I want to talk about the business, and I want to talk about enterprise tech in New York. Um, so starting with the product. So at a, at a very high level, because we have a, An audience, ah, both technical and non-technical people, um, so, and I know you don't like the term necessarily, but at a very high level, what, what, what is a NoSQL database?

A So what is a NoSQL database? Um, so I think it was, um, was it in 2009 there was the first NoSQL meetup, I think, in that summer in, in, uh, San Francisco. So someone coined that term, I think, around that meetup to give it a name. Um, so it, it's a bit of a strange term, but, uh, but I think we're kind of stuck with it for the moment. Um, we could talk about big data maybe, but If that's a little broader, so it's sort of like a subset therein. Um, I think what they mean is, ah, kind of new generation, non-relational databases or data stores. And, ah, they, so it's not about the query language, the SQL part, it's really, but they are non-relational, at least the way I define the space. And people, and it seems consistently in this space, you know, there's been a lot of products over the last few years there, um, They've all been non-relational because people are trying to scale horizontally, including on commodity hardware and commodity networks, for example, in the cloud, and sort of how do you do that and how do you make it work well, right? And distributed joins is a hard problem, so the, the solution that's been chosen by both us with Mongo and, and, and the other NoSQL products is to say, you know, we're not going to do joins, and then the question is how do you build something that can actually do something interesting without that? Right? So that's where the creativity c…

AI assessment note: “what they mean is, ah, kind of new generation, non-relational databases or data stores.”

Q a few years ago to open source a product, um, but, uh, I believe that was not the case initially, um, and as we all know, open source is a, um, it's, I guess, a strategy, if you want, that more and more companies seem to be using. So, what, what, how do you decide to open source a product, and what are the keys to making that effort successful?

A Yeah, so, so it was always open source. There's a, when we first started working on it, we were thinking about platform as a service, and having it be part of a stack, and then we focused solely on Mongo fairly quickly, but, um, but that stack was also open source, so it's kind of always been open source. Um, I, you know, what to think about, I think with open source it's sort of, um, Um, one thing I would look at is just how big is the market you're in, because there is a little, you are going to get some revenue compression, right, by being open source versus not open source if you have the same number of users, right? So if you're in a big market, that's okay. There's still plenty of room, right? And then the other thing would just be sort of, you know, is it mission critical? So databases are mission critical, so I feel like conceptually, you know, if we provide good service, people, Will be willing to pay us. But, you know, if I have something that's sort of like, ah, very discretionary, right? You know, like, um, like maybe my version control system is super important to me, but if it's down for four hours a year, you know, the world won't end. Like, no production systems crashed, for example, so do I buy commercial support for that and services? And it's like, well, I might, but it, it, it will be more optional. So I think that's the other aspect there. And you kind of l…

AI assessment note: “one thing I would look at is just how big is the market you're in”

Q So the NoSQL database segment, or world, whatever you want to call it, has seen a lot of companies and open source projects appear, so there's key value stores, there's Bigtable, there's graph databases, so Mongo falls in the document database, I believe, so can you help us understand the nuances between those categories and possibly which use case they're more applicable to?

A Yeah, I think everything in the space, um, has a goal, the way I define the space, has a goal of scaling horizontally on commodity hardware. And, um, after that, then you see some variation on a few dimensions. And one dimension is data model. So the data models tend to see our key value, Um, then what I might call a big tile, big table style data model, which is kind of, ah, um, what like an HBase would, would, would use. Um, kind of very similar to what's in the big table paper, um, that they wrote a long time ago. And then the document oriented model, which is used by some of the projects like, ah, Mongo or CouchDB and, and, and others. The, um, and then the graph databases too. I personally, I think this isn't, um, I think necessarily agreed, um, by most people, but I personally think of graph databases as a thing unto themselves. Like, this is a very interesting technology, graph databases, and if your data isn't, going back to this question of what's the original shape of your data sort of before you even start building the system, if your data is a graph, Look at graph databases. But I, I don't think if your data is not a graph, then trying to map it into graphs is not super intuitive to me, so I don't necessarily go there. So I feel like they're their own class of thing, and I would use them when I have graph data, and I would use something else when I don't. It's just …

AI assessment note: “I would use them when I have graph data, and I would use something else”

Q point? Because, I mean, those are almost, ah, different type of sales, right? One, um, especially doing startups at scale, you, you know, you can download the product, and then this minimal, you know, help you, you, you provide when you go and sell to a large account, then, you know, it's a longer a sales cycle, you have to have salespeople. How do you, how do you manage that?

A Well, I think you do all the normal things, like have a sales force and, and, and do traditional, kind of classic sales to enterprises. But, but I think that transition of the product moving into bigger companies has been very organic. Right? And, ah, They use it because, one, they've heard of it just because there's enough usage that, that there's awareness out there and they find it useful. So, uh, you know, I think from day one our intent with the design of the product was that it would be useful for organizations of all sizes and, you know, if you're building something that can scale up and hopefully building something that's robust, then in theory it should be useful for enterprises, you know. I, I think a lot of The thinking behind Mongo, you know, it, it's sort of what I wish I had at DoubleClick, and, and we were doing some kind of fairly complicated things, and, and there was scale there, and, and we could never be down, and almost never, and, and so it was, ah, um, there was a bit of a rigor to it, so it, um, that was always in the back of her head, so it makes sense to me That the product can be used in all these different sectors.

AI assessment note: “you do all the normal things, like have a sales force”

Q So assuming you have those two things, how do you go about building a community? How do you, um, make sure that it's, you know, alive, that people contribute that, you know.

A Yeah, um, That's a good question. It's a hard, it's hard, and, and we've done a lot of work on it, and I think you gotta be very focused on that if you are doing an open source product, and, you know, you, maybe, I mean, I think it's very common to just do one, and it's not even a company, you just want it to be used a lot, right? So you can imagine doing something like that too, you just, you have the same problem though, right? So you gotta get the community to be vibrant and active, and, you know, I think part of it is just being really proactive on that, with the community, getting out there as much as As possible, talking about it, meetups and things, and, but then the other part is, is, is, you gotta get people a little bit passionate about it. They gotta, you know, they have to kind of get it and, and, and to that level. You know, if somebody likes the product, they'll use it. They won't tell anyone about it. If they love it, they'll tell their friends about it. And of course, if they don't like it, they won't use it. So, we have, if everybody likes it, but nobody loves it, I think that can be a challenge.

AI assessment note: “being really proactive on that, with the community, getting out there as much as As possible”