 From New York, it's theCUBE. Covering Big Data New York City 2016. Brought to you by headline sponsors, Cisco, IBM, NVIDIA, and our ecosystem sponsors. Now, here are your hosts, Dave Vellante and Peter Burris. We're back, this is theCUBE, the worldwide leader in live tech coverage. Really excited to have your own Javier here. He is the founder and CTO of Iguazio, a company you may not know. We know, because we've been talking a lot on social, some of the smartest people on the planet, participated in our crowd chat. A lot of great perspectives. You're on, welcome to theCUBE. Thank you so much for coming up. Thanks Dave, thanks for having me. Yeah, you're welcome. So we're excited for you guys announced, basically announced the coming out party yesterday, right here. Yes, we've been here in theCUBE before, but more sort of doing architecture launch, describing a little sort of our concept. It was pretty revolutionary. So we said we'll do it in two phases, but yeah. And we got the car, this is you showing this at your booth. So tell us the story. What's going on this week? You announced your products and you've got the... So what is Iguazio? We have a bunch of great guys, founders of Xtreme IO, guys from XIV, guys from sort of different units that understand the entire stack. And everywhere from sort of the lowest layer of how flash works and disks and networks, all the way up to applications. And we basically decided that the existing stack doesn't work for new workloads, for analytics, for big data, and for the transitions that we're going to talk about in a minute. And we sort of re-architect the entire stack. And this re-architecture that allowed us to basically produce 100x better performance and things that most customers think are unbelievable. You know, our database level performance are like few million ops per second on a given server. When traditionally if you look at like Amazon or even Cassandra and some of the open source, it will be in the tens of thousands. So it does sound imaginary. And at the same time, we did build it extremely dense in a way that it allows us to actually lower the cost. So our solution is 100 times faster and 10x slower than most of traditional solutions because we basically broke the entire stack and meshed it up together. That's generally. And you've written, you write some great blogs and I guess that's the kind of thing you gotta do when you're starting the company, right? Help get the word out. But one of them really talked about the challenges that customers have with Amazon, with all the different components and the services and the variety of APIs that they have to deal with. You're trying to solve that problem, are you not? Yeah, so basically we have three major innovations. One is sort of this extreme efficiency and performance. The other one is we have what we call a multi-model database model. That means that we can basically build any type of data on top of our platform, whether it's files or streams or objects. You can actually ingest the data as a streaming element and read it as a table or scan it using an analytics primitives. And that requires a complete rearchitecture of how the engine works. It is leveraging the modern architectures because many of the existing databases were basically foreign around hard drive limitations which are very sequential by nature. So you wanna avoid in directions and things like that. And by having more flash, more memory, that means you can do more in directions, especially if you have all the things that we have in our stack for prediction and all that of IO. And then you could actually create those multi-model elements. Now having a database and running a file system on top of it buys you additional things. Like for example, you can run SQL statements to look for files. Give me all the files belonging to Joe from the last week. And we think that with this enormous growth of data in the different disciplines, IoT, big data, analytics and software as a service, you need those capabilities to handle data. So basically we have one platform that stores all kinds of data at much faster speeds than anyone else. And that also allows you to create different pipelines because instead of thinking data, streaming it in and then taking another package that basically migrates it into another form and then thinking that thing and turning it into analytics, we actually took some Amazon blog post of saying this is how you would build the stuff. And we showed that instead of six different elements that you paid, the most expensive element in Amazon actually is the TPS, not the capacity. So you actually pay six times for every transaction in the Amazon case. So beyond the 100X is actually even cheaper because you have one platform. It also simplifies the application perspective because instead of just producing sort of low level APIs, we went the Amazon way and introduced all the sort of restful APIs which we love them. That's why we sort of have those cars and all that in the show. We showed that an Android client or an Angular app can just talk to our system. It doesn't need any mediation. One of the things we did, we basically have a layer seven firewall within the data pipeline, an extremely fast one. So every transaction that goes into the system is classified. So now I can say things that come from cars are not allowed to touch this data. Things that come from Android apps are not allowed to update this data. And this is a totally new paradigm. I don't need API services. I don't need all this glue logic to ingest data. And the minute I generate an event, immediately I see that in an analytic dashboard. Another great thing we did is we took this notion of platform as a service. I think more and more people don't want to manage servers. They don't want to manage disks. They want to manage apps and data. And this is sort of what we've innovated around. There is basically focusing on data management and not on infrastructure management. So you've done some very, very interesting things with technology that allows you to do that. So by compacting, reducing the size of the data, you're able to reduce the amount of data that has to be moved, but at the same time, you can actually process it. You can act on it. You can do things with it. So talk a little bit about some of the special sauce that you've developed that makes it possible to do all these things. Yeah, so first, Dave knows that from our chats in Twitter. I'm sort of not a fan of traditional storage, like block storage, because think what they do, they take a 4k size block, and then they compress, and they do put and do all those things. And if I need to search the data, I have to decompress this 4k byte and then start figuring out where things are at. And instead, I rather do that at the database level. I rather encode the data instead of just compressing it. If I see a string saying username, I basically compact it to a number saying seven. And then I search, look for seven. I don't need to look for an entire string. I basically, I put a lot less data into the memory. I waste a lot less bandwidth on IO and all that, and I do more efficient things. You also tear the data, and not sort of basically throw everything into the block. You need to treat indexes differently. You need to treat metadata differently and blobs differently. And that means that your data is scattered across different technologies to maximize the value. Now, if you're using traditional disk technology and you have all this burgeoning and snapshots and all that, it doesn't hold water because your data is actually scattered across five different types of storage for the same transaction. You cannot gain consistency. And some of those ideas sort of, we haven't invented them. We took ideas that, you know, sort of my work with hyperscale customers, and we just evolved them because I think us coming from low level infrastructure understanding, we took the same concept that Google and Amazon have and just deployed them on a much higher performance architecture on the software. So what, when somebody buys a Guasio, what are they buying? What are they getting? So basically we're a platform as a service. It's not an individual software component. It deployed in two forms. One is you can buy the skin appliance, you know, fully integrated, sort of the Nutanix model that we do like a bunch of things in Nutanix. And another one is we'll sell you the software on a reference architecture. Our reference architecture is very sort of high end configuration. Every one of our appliances is about 400 gig ports and 24 NVMe devices. And each one of those can displace about 60 servers on Cassandra or something like that. So we basically, what we did is a very high density approach. Every server connects about 300 disks compared to a dupe approach of 12 drives per server. That is really the secret sauce that allows us to, on one end, increase the performance, but on the same time, lower the cost to serve the archiving storage kind of cost levels. So the idea is if you're, for example, we're only working right now with tier one customers versus where it's still beta level. So we like the more lucrative customers. So these are banks, cloud service providers, even Amazon, large Amazon customers, software as a service companies. One of them is sort of an Uber-like use case. And they took our stuff and they used Amazon Redshift. They took our stuff with Spark, and they got about 10x better performance at way lower cost. And that's how. Role to Amazon Redshift. Compared to Amazon Redshift. Amazon Redshift is about $1 per gig per month versus S3, which is only $0.03. So everyone likes to quote the $0.03, but they forget that when you buy Dynamo or Redshift or the more sophisticated services, it's way more expensive. So we sort of give them a price model of something which is way cheaper. We've produced 10x better performance on the application level, and they can now also own the storage. One of the challenges many organizations have, they're afraid of moving their data into the cloud, whether it's because governance issues, whether it's because they just want to have the freedom to move between one provider and another. And we have very close partnership with Equinix. They were also quoted in our press and some joint customers. Basically, we put those racks in Equinix in sort of a hub connected directly to Amazon and other service providers. And you can actually do burst computing. You can actually run VMs in Amazon and analytics on our platform. And we have a sort of dark fiver into those places. So now it cuts the cost even further because the cheapest thing in Amazon is spot instances. Just VMs without any disks. And especially if you have people that need to run VMs, one of the things we see in R&D, if I keep the VMs, sometimes I ask for my VMs because I'm also sort of a geek and like to code. And they tell me 100 bucks a day because I have to keep the VM. But I'm not really using it. I need to write blogs and do other things. Build cars. Yeah, build cars. So really what you want to do is when you turn off the VM, you want it to sort of disappear. You only want to pay a dollar for the time you were using it. So you want to decouple the storage from the computation and just launch. And the essential thing is basically having the separation that the VM doesn't hold any state, which is this sort of cloud native architecture. So your licensing model is a subscription? It's an annual subscription sort of model and determined by two sort of main factor, performance and capacity. Whether that's the same if I buy an appliance? Yes, basically when you buy an appliance, we give you the hardware at cost plus sort of movement fees and that sort of thing. But what we want to make it very transparent to users, we're not after being sort of an EMC model, 60, 70 points margins across hardware. We're basically a software company. But what we see is that customers like this notion of an integrated turnkey solution. So we say, okay, we'll give you this integrated turnkey, but we'll charge you for the software. In part, we're working with all the old regular suspects on sort of the OEM side and potentially have others resale a configuration with our software as well. And they would buy perhaps the software of the reference architecture, an SDK or? Yes, and so we have a lot of experience with it in the past, both in Voltaire and Manux where I was, we were always in sort of an OEM model with most of the, you know, San, Oracle, IBM, HP, et cetera, where we were doing the sort of the customer sales, but the fulfillment was is OEM model. So we're quite accustomed to this kind of model. We're still early stage, so we're working directly with customers. But long run, it's sort of a partnership game. So what's your take on, I mean, you saw our server sand report, you see the sum of the work that we do. What's your take on hyper-converged infrastructure? You said you mentioned Nutanix before as a model. Yeah, I think they're a great company. So I don't have any reason to bash them as a company. I think in general, the HCI model is not a long-term sustainable model because what the cloud guys, you know, basically what you see in the enterprise side, and I don't, it's not just Nutanix. It's all this sort of HCI movement, et cetera. They're basically trying to create an Amazon EC2. That's something Amazon did five years ago, okay? And if you are going up to the cloud and looking, where is, you know, Azure investing their money and where is Amazon investing their money? Where is Google investing their money? Is definitely not there. Is providing platform as a service, you know, buying AI companies, you know, building services for doing querying and analytics and all that because customers don't want to manage virtual machines. They're sort of joking, joking in one of the tweets. It's sort of the wide generation, you know, they want things to work. They don't really want to do what we used to do, you know, bring up VMs. They want the cake, not the bake. Yes, and if you're looking at segmentation of the market, you're going to have companies with like hundred folks, okay? They don't need any infrastructure. They need Office 365. They need the RP in the cloud. They need apps. They don't need any infrastructure. And if you're looking into bigger organizations where you're going to see the big movement and it's not going to change overnight, you know, people are still using mainframes, is basically the move to data in the center, I call it. Basically, the organization will try and centralize more and more data, create governance around it, and the compute is going to be the floating point, whether it's going to be mobile devices of the workers, whether it's, you know, the website that may be even hosted somewhere. So it's going to change the paradigm. And in this environment, HCI is not the right play. So HCI is the right play for the existing architecture of people that don't like sands and don't like fiber channel and all that stuff and still play with LUNs and VMware and Oracles. But if someone is building a cloud native architecture or a data-centric architecture, they're either going to go to a cloud or they're going to create an a cloud-like experience for the enterprise. So again, it's not going to go away, but what I had my comments on your report, it was sort of so biased towards the entire world is going to be sort of HCI. And I see that from our discussions with CSOs and CIOs and all that, that's not where they want to take the company. They want to move to pass. And I think there is sort of, if you look at sort of the CIO and IT folks, there is some tension right now because sort of IT, what they know is VMware and SAN and Word by names and all that stuff, okay? And they're sort of having a hard time evolving into this sort of service model. Some of them are evolving and they actually own the relation with Amazon, for example. So they're no longer the guys that making the thing work. They're also the brokers for the service. But what you would see is that if they want to evolve fast enough, the business units, the guys that serve are the shadow IT that are basically taking up the wallop and doing the development in Amazon because it's sort of too difficult to do it on-prem. They'll either go to Amazon or say we want the Amazon-like experience on-prem. So I think this movement will be faster than what you guys are anticipating. And that's a reasonable look. We've had a fair amount of debates inside about what's going to be the local, the focal point or the low side of where the value proposition is. Is it going to be IAAS? Is it going to be PASS? Is one going to supersede? I have my way of thinking, but these are good. This is great feedback. One question before we go, if I can? Yeah, we're out of time, but go ahead. Oh, okay. Okay. It was quick. Next time you're on theCUBE, let's talk about whether or not developers, how much developers need to know about the platform and how better performance makes you a developer. Yeah, we're trying to make it totally transparent. And we actually have, I have in a couple of hours a meetup in Docker in New York and we're going to talk about serverless and API-driven computing, complete cloud native. We're actually going to show some demos. But basically we're going after the big megatrends of IoT, which is very data-driven. Cloud native apps and SaaS providers. Also a lot of very developer-driven, stateless and all that. And the third one is sort of the big data camp with sort of next generation kind of data warehousing and analytics. So these are the three things that we, as a company, go after. I think those are sort of the biggest megatrends. That's why it's a lot of fun. So last point, tell us about the name and of course the logo. Tell people where that came from. So the name is, there is Iguazu Falls in Brazilian Argentina junction, which is sort of, I think, the prettiest waterfalls in the world and also very massive. So what we wanted to convey is the message that storage used to be really static and now it's sort of huge volumes that sort of flow immensely, coming from many different sources. If you know that, they have about 100 different rapids. And so that's sort of the source of the name. Excellent. All right, we got the content flowing wall-to-wall coverage here. They did it. That was furrier like. Furrier like, thank you, I think. Keep it right there, everybody, we'll be back. This is theCUBE, we're live from New York City. Thanks. Thank you.