 Live from the Moscone Center in San Francisco, California. It's The Cube at AWS Summit 2015. Live here in San Francisco, I got a little periscope action going here on my phone here at Amazon. Web Services Periscope is a new Twitter app. I'm John Furrier with SiliconANGLE. It's The Cube. Mark Farley, my co-host. Hey, now. Noam Shindar, the Chief Operating Officer of Zadara Software. We're live, we're periscoping. And we got The Cube with The Cube team here. Welcome to The Cube. Good to see you again. Thank you. Good to see you too. Tell us about your company, what's going on. I'll see you guys are doing some great stuff right now. You came into our Palo Alto office to talk. What's going on here at Amazon Web Services and what's the tech angle? It's deep tech here. It's not a big marketing show. It's all about technology. What's going on? Yeah, we love this show because people are here to buy. So they have real problems that need to be solved and if we can solve their problem, they buy. And if we can't, they move on to the next vendor. That's a good show. I love that. No kicking tires, getting business done. And what we do is high-end storage services for Amazon customers. And we do this both at Amazon in multiple regions and we do it on-premises too, which makes us unique because it means hybrid services from Amazon or to Amazon. We do this with file services. We do this with block services. And we do it using the same business model as Amazon. Buy the hour, elastic up and down. And yet it's on dedicated hardware. That's the magic sauce. Dedicated hardware on demand. Yeah, so that's real interesting, right? I was looking at your website and seeing on-premises as a service. Yes. It has a headscratcher for me, right? I mean, it's an interesting play on words, but how does that work? You're charging by the hour or for hardware. That's right. Really? Yes. And this is a business model that works. Can you please explain? Yes, it works wonderfully. So start with what we do. What we do is we instantiate storage clouds in many locations around the world, and we manage them from a very small number of locations, which means we're very efficient at building storage clouds. We can not just run them, but run them efficiently and run them with extremely high reliability anywhere in the world, which one day led to the realization that we don't have to do this only in public clouds. We can do this in private ones as well, just like we can bring up our 33rd public location. We can bring up a 33rd private location. And the other thing to understand is that this is designed to be very, very cost-effective, which means we're able to deliver this to you with not a lot of cost to us in terms of hardware, because we run on top of commodity hardware. So you, as the customer, get with no upfront fees, so for zero, for free, you get storage on your premises, and we charge you for your actual usage of that using that hourly model. And the reason that works is twofold. One is it's very attractive and people grow their usage over time. And the other is we do ask for a six-month commitment, which really is to keep our logistics under control. No, that's reasonable, right? I mean, you really don't want somebody experimenting with this for a week or a month or something like that. You really do want some amount of commitment there. The management model for this sounds really interesting, right? So it sounds like you have some sort of centralized management of these. But customers want to do that, too. Do you share management responsibility, or how does that work? Yeah, we do. We share it, and we share it roughly along the line of low-value, annoying, menial activities. Those go to us. And high-value activities go to the customer. So in other words, the responsibility for firmware upgrades. We actually don't have firmware at software, but the equivalent with the traditional providers, like the person who was sitting in the seat before, they have firmware, so we do the equivalent. We do software upgrades. We're responsible for those. We do them transparently to the customer. We do tell them that it's happening, but there's nothing that they need to do. Things like that. Responsibility for hardware replacement. Drives die over time. It's up to us to send new drives to replace them once it failed. Those are the low-value, annoying, menial activities. That's on us. So do you have a dual-controller architecture then? We have a self-healing dual-controller per virtual array architecture. So it means two things. On the per array, it means that, unlike the traditional storage architectures, you can spawn as many virtual arrays as you need, and they can be as so small as to be a single application array each. So you can have as many of those as you need, and in terms of the controllers, it's two, but it's self-healing because they're virtual. So should a controller fail, for any reason hardware or software related, we'll simply spawn a new second controller. So you always go back to two. Interesting, so it runs on Intel hardware, and you just bring up a new virtual machine, which is your controller. So when you do firmware upgrades, then are you just killing one of the old controller machines and instantiating a new one? Yes, and we're doing a failover. So we're doing an HA failover from version minus one to version zero, from version zero to version plus one. It's that simple. That also allows us to do other neat things. First, it tells you that we can be heterogeneous. We don't have to have the same version running in the entire cloud. We can actually mix and match versions. But also means we can grow and shrink the controllers. So let's say your controller's too small, well no problem, you push a button and it will fail over to a bigger controller, and then your bottleneck is gone. That's slick, that's really slick. So Nelson, now Nelson's the CEO. He is, yes, co-founder, CEO. Did he invent this? Yes. Yeah, he's a very clever guy. Yes, the day that he told me about this is the day that I knew I want to work for this company. Oh, so did you work for the company before or after you found out about this? I'm the first employee to join the company who's not a founder. Ah, okay. That's like being a founder though. When you're in a real stage company, there's the founders who have the idea and then there's the sub-founders who come in, with the founders, who are the first high, basically technically first-highers. So that's, you know, to me, it's all entrepreneurship. When you hit like 15 employees, that's kind of wave two. When you're in the sub-15 number, it's all craziness. Exactly, exactly. I mean, mop on the floors together, you're doing all kinds of odd jobs, you know. It's total craziness, you're trying to get some traction, you're working hard. So I got to ask you, so Amazon talked today about recycling compute and all this stuff, this agile-like, when we talked in Palo Alto, you were talking a little bit about this usage model. Can you expand on that? Is it the same as what they're offering? Because you have a unique economic value proposition. So one, share that with the audience, then compare that with what Amazon offers. So you'll have to bring me up because I actually didn't sit in the keynote, too much business happening. I'll just tell you about yours, and I'll bring up my notes. To the earlier point, too much business happening on the show floor. How many deals did you close? I think that we, just in that time of the keynote, I think two new customers. Fantastic. Yeah, just in one keynote. How much revenue you guys are doing? We can't disclose revenue, but we're in the millions. Okay, so I had to ask, sometimes it comes out, just cubed, just falls out on the cube. Trust me, I am dying to tell you. You're smiling, you got a spring to your step. Yes, absolutely. This is good, so what is it about your offering that aligns well, your closing deals, what is it about your product that's winning, why are you winning? So at AWS, we're winning because we're providing storage that is solving the problems of the most demanding customers. If you think of the pyramid of customers, Amazon does a fantastic job of covering the wide base of the pyramid and up to a certain level, but there's the tip of the pyramid that have very specific, very difficult needs, and we solve them. It could be unavailability, it could be on features, it could be on replication to on-premises. So that's one reason people choose us, another reason people choose us is support, 24 by seven support is included in our product. So the enterprises we use us don't like to read the manual or have to search for help. No one reads the manual, RTFM, everyone knows what that means. Guys that use Microsoft read the manual, they have to. So we're there with them along the way. Anytime they need us, we help them. In fact, we insist on speaking with them when we begin, just to make sure that we can cover some important things with them. Here's the problem I have with storage right now. I have way too much capacity and I'm paying for stuff I haven't used. What are you guys doing in that area? With us, whether it's on-prem or in Amazon, you never pay for capacity you don't use, because it's elastic by the hour. So there's no reason to ever over-provision because you can just add it. In fact, we have a feature called Hotspair, which is you can have an extra drive that swaps in in case of drive dies. None of our customers use it. We have exactly zero customers that use Hotspair. We're not losing any revenue, it's a free feature. Why do customers not use Hotspair? Because the alternative to Hotspair is should a drive fail, you go into our interface, you buy a new drive, you swap the dead one for the live one and then give back the dead one. That's one minute of clicking around and it's done. So why pay for a drive you're not using if you can just under a minute swap in a new drive? So John, to your point, with us there's never a reason to over-provision because more capacity is at your fingertips any time you need it. And by the way, should you reduce your usage, you can give capacity back and stop paying for it. We have a customer who just did that. He said, my CIO told me I have to reduce my overall IT spend. Can you guys help us? He said, of course. Just delete some things, consolidate onto a smaller capacity and then give us back the capacity you're not using. And we'll resell that capacity to somebody else. So that's obviously cloud capacity, it's not on-prem. Right, but on-prem, we charge, on-prem, same story, we can't recycle the capacity. It's hard to recycle out there, yeah. We can recycle it at the node level. A node is about 200 terabytes. So if somebody on-prem stops using a node, we could potentially, we own the node. So we can take it back and deploy it at another customer. But as far, but that doesn't affect the customer. Customer only pays for what they're using. So the same hourly business bottle that we have at Amazon and the same price that we have at Amazon is also what happens on-prem. So if you're using 50 terabytes on-prem with us, you pay for 50, you grow to 60, you pay for 60, you shrink to 40, you pay for 40. So EBS is now advocating basically SSD. Yes. You're on-premises nodes, how do you allow somebody to mix and match between SSD and spinning rust? We provide any option the customer wants from all flash array to all magnetic or anywhere in between. The customer chooses the mix, unlike everybody else, they can change the mix. So one of our mottoes, if you will, is you can't make a mistake. Because every choice you make with us is reversible. So you bought too much flash, no problem. Give the flash back and get some magnetic drives. Not enough flash, no problem. Get some flash. By the way, we use flash two ways. You can use flash's capacity. So for example, an all-flash array. And or you could use flash as a cache. So standard included in the price with every magnetic VPSI, virtual private storage array, is an SSD read-write cache. And you get, depending on the size of your controller, that determines how much flash cache you get. And then you always have the ability to expand that flash cache at the push of a button non-disruptible. So if your data set grew and no longer fits in cache, no problem. Two clicks and you're done. So I'm curious how functionality like replication works. You can replicate from one site to another site. Customers like to do that, but I think customers are also looking at being able to replicate from on-premises to the cloud with the idea of doing recovery in the cloud. Is that something that you folks offer too? Oh, absolutely. It's one of our killer features and almost unique. So probably the only other company that provides the same storage on-prem and in the cloud. Because it's a virtual controller. Exactly. So you're running a virtual controller in an EC2 instance? Actually, we don't. We put our, remember, it's dedicated hardware on-demand. So if we run an EC2, it becomes shared, not dedicated. So we place physical infrastructure next to Amazon, but using an Amazon business model. So you get the best of dedicated hardware in the sense that you're not sharing, but also the best of cloud because the business model is cloud. It's completely on-demand. So on-prem to the cloud is we will deliver storage to your on-prem and we already have the storage at Amazon and Azure and other service providers. And you can replicate in either direction. It's the same storage exactly. You don't have two products. We're a single product company deploying the storage many different places and ways. And you're able to replicate not just from one location to another, but also one to many. And you have very nifty features. For example, you can do a failover for disaster recovery, but also fail back. What a lot of vendors miss is the customer doesn't want to stay in DR mode. The customer wants to revert back to the original as soon as possible. But we make it really easy to do that fail back because our system will recognize when the mirror broke and then we'll sync only the data from the point the mirror broke in on back to the primary site. So tell me about the vision you guys have and compare that to you heard from NetApp. They're an incumbent, but they're renovated with Amazon. You got EMC transforming. They're a big aircraft carrier. They're moving in a direction. You've done a great job over there. You've got Pure Storage, a new guy in the block, kicking some ass with their number, so to speak, all flash arrays. You got HP Storage, which we'll get the update HP Discover. You were previously there, so you should try to give some commentary on that. IBM, they have a storage group. In fact, the April Fool's Day joke was... Yes, that was funny. It was a piece. Dave Vellante felt funny. Dave Vellante. I know people. I know people. You know people buying their assets. Actually, people believed it. So that's how interesting that market is. You have a massive shift happening. At the same time, Amazon's pointing at an inflection point. So two major forces going on in the market. Technical change, business economics, that all the theaters are exploding in disruption and innovation. What's wrong with people out there on the old side that aren't moving fast enough? What are they doing wrong? And what are the guys on the old side doing right to get into the new side of the street, if you will? The business model is wrong. It's everybody who's saying that they're doing cloud storage but isn't providing it as a service business model with consumption pricing. And I'm not talking about the so-called flex leases. I'm talking about true consumption pricing. By the hour, if I shrink, I pay less. If I grow, I pay more. So what everybody else is doing wrong, are they do the capex model or long-term leases or these flex leases that have a big base capacity and the ability only to grow past the base capacity? Customer wants the ability to shrink. Bad stuff happens. We've all lived through down cycles. So you're saying the line in the sand between the two sides of the streets, old school, no school, is consumption. Yeah, consumption price. Is there anything else you'd add in there? Elasticity? Well, consumption pricing and elasticity are together. If you're inelastic, then you're not, you don't have consumption pricing. Because consumption pricing paying from, exactly, it's built in. What about something like warranty? Do you have warranty in consumption pricing or does warranty just fly out the window? So this is why we call ourselves, and it's a great question, we call it sort of enterprise storage as a service. What's wrapped into as a service is consumption pricing and an SLA, better than a warranty. The warranty just means if something breaks, we'll replace it. Break fix, right. We say the thing will keep on running optimally or you don't pay us. So not only will we fix it, we actually won't have hiccups. We won't have downtime. That's our promise. So think of warranty as this good and an SLA being this good. Assuming that the SLA has real promises in it. You can tell, if you're examining an SLA, just look at the penalties. What happens if you don't live up to the SLA? And that's- Yeah, that shifts the responsibility for whatever you're running on from the consumer to the vendor. We have some of the biggest companies on the planet putting out our FPs saying we need a lot of storage. We don't want to deal with it. We need what it does. We don't want the headaches that come along with it. So we'll announce relevant deals as our customers let us make them public. But think about traditional manufacturers, bricks and mortar, very, very conservative, saying we will not buy any more storage. Yeah, so one of the big drivers of the storage industry for decades has been the fact that the equipment all becomes obsolete three and a half, four years, right? It goes off maintenance. And once it goes off maintenance, it's too expensive to pay the maintenance for it. If they're not paying maintenance, it's a whole different world. And if you're using SSDs, it's a whole different world. Right, and so because of our architecture, we can swap the underlying hardware non-disruptively. So the era of migrations is over as far as we're concerned. The customer can keep running and because we're the service provider, even on premises, we can completely refresh the hardware non-disruptively to the customer. And a magical thing happens when we do that. The price goes down. So one day you pay this much for storage. The next day, the price goes down. You didn't have to ask for it. You didn't have to scream and shout about it. It just happens. So why is all of this happening? I want that to happen every day. So this, well, Amazon is forcing the world this way because Amazon doesn't go down when they change the software, Amazon doesn't go down when they change the hardware and the prices go down of their own volition, right? So the customers are starting to say to their traditional providers, this is the experience I get at Amazon. Why can't I have this on premises? And the vendors hammer and haul and say, oh, it's revenue recognition, it's my model and I can't change it. Well, the cost per terabyte or whatever is always decreasing, right? So if you buy new capacity, you're getting it cheaper than you bought it three years ago or last year or whatever. Yeah, but what I'm talking about, my old capacity goes down in cost. How's that? Hey, Nova, I got to ask you a final question here. Okay, we've got good data on what you guys are up to, how it relates to the trends here. What outside of storage is impacting your business? We heard from Amazon here about the container service, little Docker actions to application agility. You've got Lambda for mobile. They're putting everything in the stack. Storage, obviously the Bakken object. What outside of storage is putting pressure on your innovation strategy? It's customers looking at services broadly. So when we started, it was a point interaction. I have this one thing. You don't mean professional services, right? Sorry? As a service, yeah. So X as a service, everything as a service. Is the, so our customer interactions are shifting from point problems. They have this project, this application. Can we help with that? Okay, thank you very much, Don, we're happy. To, we're re-architecting everything to be agile, flexible, elastic. Can Zadara be at the heart of that? So it's forcing us to also think at that level as far as how do we integrate with compute? How do we integrate with different cloud platforms in terms of automation, orchestration, containers? All of those things are now entering through the conversation. Whereas before, we, you know, we were- Well because they're software based and their application workload driven. So you're saying the applications are dictating to the network and infrastructure, versus the other way around. Yeah, storage forever will be a necessary evil. And I'm okay with that. But nobody wants storage. I have an application and it has data and that data has to sit somewhere. But I don't like storage. I don't, you know, I need it, but I'd rather not do it. Kind of making me tear up. Yeah. Yeah. But- He's a storage guy. He writes books about it. Me too. He loves storage. It's not storage to him. But this is, I have- Storage is sexy now. Come on. Yeah, again, I've seen the headlines. It's still bits in a bucket, right? So yes, we make it easy and we make it fun. But at the end of the day, it's driving an application. The application is what the customer cares about. And the storage has to enable application and keep it running optimally. So we always think about what's the customer doing at the end of the day and how can we support that? Noam Shendara, who's the chief operating office of Zadara Storage. Cutting deals and closing business at Amazon. Congratulations of your success. And, you know, again, the line in the sand has pretty much been laid down. Insumption, I mean, being elastic is inherent in that model. Congratulations. And the world's changing in storage. It's certainly infrastructure. It's theCUBE. We're breaking it down for you here, live in San Francisco at Amazon Web Services Summit. This is theCUBE. We'll be right back.