 Live from the Mandalay Bay Convention Center in Las Vegas. It's theCUBE, covering VMworld 2016. Brought to you by VMware and its ecosystem sponsors. Welcome back to VMworld 2016. This is theCUBE. We call it our Iron Man, going through three days of coverage here. I'm Stu Miniman, join my co-host Mark Farley, storage pro, welcoming back to the program. Two guys have been on theCUBE many times. We actually helped us with some pre-VMworld activity from our East Coast studio, Randy Arsenault and Steve Kenneson, both are from Infinidat. Randy's the CMO and Steve is the Vice President of Product Marketing. Guys, thanks for joining us. Thanks for having us. Thanks Stu, appreciate it, good to be here. All right, so first of all, how's the show been going for you? It's been fantastic actually. We've been doing a lot of comparisons of our experience this year versus last year and everything has just scaled up like dramatically. The activity we've gotten in the booth, we increased our sponsorship level, so we've got a bigger space. We just had a customer presentation this morning. Little teaser, the customer's going to be presenting with us in the next segment, which is really awesome. That was well attended. The session ran a half hour over because we had so many questions. We've had a ton of analyst briefings that have all gone exceptionally well. So I think the level of visibility we're getting and the fact that we're kind of aggressively positioning relative to the all flash array vendors who are kind of dominating the show has gotten us a lot of eyeballs and a lot of attention, so I'm thrilled. Yeah, I mean, anybody that's been to VMworld know there's always more here and more people, more things, more technology, more parties than you could ever go to. I did finally get to like, you know, sprint through the show floor and you're right. You go to any of these, you know, not just the storage shows, but everything else, that drumbeat of, you know, flash, flash, flash, flash is now cheaper, flash is better, you know. It is both a dessert and a floor cleaner. So, you know, flash is there. You guys have a slightly, you know, different look at, you know, kind of flash and disc, you know, a contrarian view, if you will. You know, why is that and how does it fit? Yeah, so our theme this year in the booth was faster than flash and easy on cash. And we had some technical presentations that were talking about how our system can actually beat all flash systems with real world workloads. So you hear this all flash performance and really the numbers tend to get inflated just because you're reading a lot of data. But if you have to do reads and writes and real world workloads, we have a solution the way our system was architected that actually has better performance than all flash arrays. So when you start taking three terabytes of DRAM, 86 terabytes of SSD and 480 spinning, you know, hard drives, you can fill a really nice system and have that perform very well for you and have capacity to grow and expand and really service your enterprise. All right, so, you know, Wikibon has not been shy talking about how flash is going to change architectures. David Fleurer, our CTO has talked a lot about how it's the things like, you know, de-duping impression and we can do that on flash. We usually don't do that on disc. And that's been kind of fundamental as some of the economics there. So is there a slight adjustment we need to have in our thinking or, you know, there's a fairly dramatic adjustment that has to happen, well, maybe not dramatic, but so, as you know, we've been speaking with David and we tremendously respect his opinion. He's a true, you know, thought leader in the industry. So we always look to David for kind of guidance and vetting of the message. So we've been talking to David a lot. We briefed him again yesterday. I think his core assumptions are valid. I mean, eventually, inevitably solid state media is going to reach the inflection point where it becomes, you know, where there's economic parity with spinning disk and spinning disk no longer becomes a viable medium. The issue or the challenge is that most of the all flash vendors bet about five years ago, most of them, some a little before that, they kind of placed a highly leveraged bet that by this time, spinning disk was going to be obviated to the scrap heap. However, what subsequently happened in the interim is number one, the elasticity in the pricing hasn't happened at the rate they predicted and the spinning drive manufacturers haven't slowed down. They haven't stopped innovating. They're constantly driving up aerial density. They're driving down power consumption. They're improving reliability. Disks stopped getting faster in 1992, right? So you're not going to get a fast system by building it out of spinning disk. However, flash is a tremendous caching layer. It's a wonderful caching layer. So use it for the purpose for which it is intended and keep your data for long-term persistent storage on good old reliable, cheap, stable, predictable medium. So as Steve said, we've architected the solution not from the media perspective. It's not media-centric. It's an architectural-centric view. It's a software product that is designed to exploit whatever media is relevant and economically viable at any given time, whatever that happens today. I have to jump in too, Stu. So yesterday, I heard one of our SEs went to Chad's actual presentation at EMC, and he had this chart where SSD was crossing over to a price inflection point for hard drives, spinning drives. And when he got close enough and he looked at the chart, it was the 15,000 RPM drives, and you're like, well, of course, right? But that's not the whole objective, and that's not what customers are looking to do, right? The ability to, again, through a special architecture, still provide the performance leveraging and taking advantage of the SSDs, but then giving customers the ability to have all of that data and have it present. All the workloads these days, they're changing, right? If I want to do big data, if I want to do analytics, I want to have that data online. I don't know what's important, so I want to be able to store everything. How do I know, right? And because someday, what might not be relevant today might be very relevant tomorrow. I want to be able to pull that up, do processing, and drive my business, and that's what we're finding our customers are doing. And because it's a pluggable architecture, and again, it was designed from the ground up to exploit high aerial density spinning drives as a persistence layer. But if we get to the point where the economics make sense and are viable, we can switch it. It's a software product that's running on commodity hardware. So it's not a huge gargantuan effort to make the switch. So are there any verticals that you're finding particularly good for you? Yes, and it's kind of the usual suspects, right? So one of the things that I find fascinating about Moshe, I mean, so Moshe and I as the CEO and founder of Infinitaz, I think most people know, this is his third major project. Obviously, Symmetrax being the first, XIV being the second, this being the third. And one of the things that I most respect about Moshe among many is that he's never tethered to the past. So whenever he looks at a problem, he always looks at it with a fresh perspective. He never ties himself to the architectural decisions or patterns of the past. He's ready to look ahead. Same thing with Infinitaz. So he recognized that flash is a really, really good caching layer. That's the purpose for which it's best suited. Much more so than a primary data store. He saw all the all flash guys who were basically the same people that nine, 10 years ago were telling everybody to put all their data on 15,000 RPM drive. So always put your data on the absolute most expensive media available at any given time because their architectures were never fundamentally altered to exploit the capacities and capabilities of that media. So they had to constantly keep cranking up the ratio of solid state media to spinning media just to sustain real world enterprise performance levels. So the approach Moshe took was, no, we're not going to do that. We're going to build an architecture. We're going to build a modular software framework that will exploit those drives in the means or in the manner for which they're best suited, but still provide the same performance. So we do benchmarks all the time. We do competitive situations all the time where, as Steve said, real workloads actually write data sometimes too. They don't just read it. So we get into these situations where we're delivering 99 or 100% of the performance profile of an all flash array, but in a general purpose, functionally rich, mainframe available, mainframe reliable solution. So at petabyte scale, so three petabytes in a floor tile, pre compression. So it becomes a pretty, and in fact, the customer who's coming up after is going to share some experiences when they did a bake off in a comparison between their legacy provider, Infinidat and all flash solutions. And the results are pretty interesting. So how many nines do you guys talk about when you talk about reliability? It's a big number. It's a lot of nines. It's a lot of nines. Seven nines reliability. And actually, it's pretty interesting. We had a Blue Lock, one of our cloud service provider customers in our booth yesterday, because we have customers who are in our booth giving presentations. And he had a quote on his slide that said, all the nines I can handle. It was pretty good. I mean, that's a bold statement because I mean, if you analyze that, how many minutes or what during a year is that? It's three seconds of unplanned downtime per year. Three seconds of unplanned downtime a year. That means that only one customer could have a problem and you wouldn't be able to say it's seven nines anymore. Correct. I mean, that's a really bold statement. I don't know if people understand that, right? It's amazingly bold to say seven. Yeah, I mean, so the benchmark for the longest time are kind of the table stakes was sort of five nines. Everybody said five nines data center, right? When you think that seven nines is two orders of magnitude, it's not like 10% more. It's a thousand percent more. And there's a bunch of different ways we can achieve it. There's a fairly circumspect conversation about the architecture, but Moshe, again, having done this a few times and generally been pretty successful at it, understood that if we're going to be servicing petabyte scale workloads, another thing Moshe famously says is petabytes are the new terabytes, right? It used to be five years ago, if you had a petabyte class data center, you were in the vanguard. I mean, you were very much at the high end of the market. Now it's commonplace, right? It's not just the Apex big data providers, it's everybody, even the mid-market guys. The puck was moving there and now there's a whole game going on. Right, so he realized that if you're going to deliver a petabyte scale solution that's unified in a single pre-integrated delivery mechanism or delivery system, you're never going to overcome people's concerns about blast radius and reliability if you don't provide mainframe class absolutely bulletproof reliability. So the first thing he did was invest in a massive QA lab in our development center in Hursley in Israel. We have, at any given time, a hundred petabytes of storage in our lab that is being used for integration testing, performance testing, software testing. Every customer system is burned in for a minimum of a month before it ships to a customer. It's adaptive, it's self-healing, we can lose, it's everything top to bottom is N plus two, active, active, active, redundant. So it's a bulletproof system. And another thing Moshe famously says is all your data's important. Like we don't want you to have to pick your favorite child, right? Because Steve said this earlier, you might have a data element that today is of little value. The perfect example is genomics research and pharma, right? So you have a disease vector or an outbreak of some weird little regional disease that gets controlled and then all of the research data associated with that is archived off. Well, when something similar pops up 10 years from now, you need to quickly be able to go get that data set again and act upon it. So the only way you can do that is if you've got a solution that makes it immediately available at that scale. So I mean there's no doubt that Moshe has huge shops and storage. I mean, he's probably the father of enterprise storage as it was. You talk about that reliability. Here at the show we're talking a lot about what's going to happen with cloud? What's going to happen with some of this application modernization? How does that fit into your customer engagements? Well, it's really interesting, right? The more you hear about cloud, everybody kind of just looks, almost looks up and says, it's just there, right? But that data actually has to go someplace. A big part of our install base, and this goes to answering Mark's earlier question, is our cloud service providers, right? We have a lot of cloud service providers worldwide. We just released one of our customer testimonial videos from Nextra data in India, right? And what do you do when you have 1.7 billion people that need access to their data like that? It's reliability, it's scale, it's performance, you need to be able to handle all that. So while yes, cloud has been a really big topic here at the show, it's got to go someplace. And what they like about the models or the solution that we sell is not only can they, I like what Randy said, as far as you don't have to pick your favorite child, your data can kind of live where it needs to live. You can have that back end scale and you can have the performance required so that the active data that you need, you have it. In addition, the way we sell the solution from a capacity on demand model, we'll roll in an entire InfiniBox, right? It's just easier, it's smarter, it's easier for the customer to worry about, sell them 150, 250 terabytes and then let them, and we'll just monitor it and just charge them as they start to use it. It's not like you have to roll in, have an SE come in, plug something in, it's all there, it's all ready for them to go. And the service providers love it because that's how they make their money. They can't really pay until they're billing people, right? Yeah, and it also creates some interesting delivery model options for us, right? So for instance, we talk about Genomics the company was sort of founded in the genomic space and we've since expanded out into a lot of other verticals and use cases, but the big challenge in the genomic space was compute is very inexpensive now, so they can do massive, thousands of core analysis exercises in the cloud. The problem is the research data sets and the input data sets are also getting bigger and bigger and bigger. So while the compute, the elastic compute is very inexpensive, moving that volume of data in and out and around the cloud is incredibly expensive. So if you can have a solution that enables you to deliver petabyte scale storage in a sort of cloud adjacent model or sort of a hybrid on-prem model, that makes those exercises now much more economically viable. So it's good for both the end users and the providers. So in that case, do you have service providers that have Amazon Direct Connect or the Microsoft Evelyn to be able to tie into it? Stay tuned. Ask Brian that question in the next segment. So Steve, when you describe kind of those 1.7 billion users, from their mind, it's a storage list environment. And from the service provider even, what's the experience for them? Do they have to think about storage or? We said storage list doesn't mean that there's no storages that we kind of move away from it. What is kind of, you think of the traditional storage admin role or the virtualization admin at HearStore, what is that transformation when they choose an Infinidot? Yeah, it's been great. The Nexra data is a division of Airtel and Airtel is the third largest cell provider, I think, in all of India, I think is what they described to us. They bought three systems. So from a high availability perspective, not only do they have high availability in the system, but they also bought one for two of their other data centers. They rolled the systems in. I think the customer experience and the customer quote was, we installed the system, it took them about four hours and by 24 hours later, they were already customers using the system. The story they told me, the experience. And when I say using, I mean, they had already provisioned all the storage out that they needed to for a few of their different customers. Customers were already putting data on the box. One of their customers is like what we have here in the United States, Fandango, the movie theater ticket thing, to Randy's point, when you get a spike in a workload, right, you can provision, compute pretty easily. So when the weekend comes and everybody's going to the movies, there'd be a spike in their need to be able to service tickets for people that wanted to buy tickets. The customer said from servicing that and the data, they found a 10x improvement in performance. So rolled in really quickly, up and running in 48 hours and better performance. And the other thing is when you think about, you know, the server, the storage list cloud, you know, which is an interesting metaphor, but the other problem is that we talk about flash versus spinning disk. Every major Apex cloud provider uses spinning disk because it's not economically feasible to do it any other way. And furthermore, the reason they're using spinning disk is number one, because of the economics. And number two, because in order to guarantee data protection using the distributed schemes they're keeping five, six, seven copies of the data. So as the volumes continue to ramp up, that sprawl becomes a real issue. Like this is a non-trivial computer science problem, right, so at that scale, being able to protect the data, we have a 76% efficiency with our infinite algorithm. So we can deliver again, almost three petabytes in a single system uncompressed and more on that, at seven nines of reliability with a 78% efficiency of usable over raw. So economically, it just makes a lot of sense if you're in a highly dynamic rapid growth environment like a cloud provider. Is that data protection algorithm? Is that intellectual property? Is it? Yeah, the solution we have at present, there's 65 granted patents that are wrapped into the OS. We have another 70 in the pipeline. If you think about it, what's sort of one of the many things unique about our company is that we've got three generations of development and design expertise. We've got kind of the old timers from the Symmetrix era. These are the guys who developed distributed RAID, the guy who wrote the spec for iSCSI. I mean, these are the titans of the industry that are kind of the core. Then we've got some of the people from the XIV project who are the next generation distributed computing thinkers. And then we've got the sort of next third generation of the Infinidad people, the younger millennial kind of developers. And it's an amazing ecosystem of intellect. So these guys all learn from each other. And when one of the young developers wants to go solve some complicated problem, he walks down the hall and knocks on the door of Chaim Kopelevich and says, how do I do this? And he goes, here, we did that 25 years ago. Here you go. So it's remarkable, right? So having that kind of expertise in house gives us the ability to have such an IP rich solution. Because these guys have already solved all the really hard problems. So now they're into the more nuanced stuff, right? So that's how we're able to develop such an elegant solution that scales so well and delivers that kind of reliability. Sounds like it'd be a fun office to work in. So our primary, we have three development centers. The primary one is in Herzli in Israel. We have one in Boston, where our mainframe development is. And then we have kind of a advanced research team in the valley. So, you know, we're over in Israel a lot. And the environment is just, it's so energetic. It's like you walk in the door and you just start to float off the ground. These guys are all yelling at each other. And it's in a very... And they drink a lot of coffee. They drink a lot of coffee, probably far too much. But they get so much stuff done that we encourage them to drink as much coffee as they want. Hey, so I don't want to end this without talking about how your business is doing, right? Business is going great. Again, thinking about where we are this year relative to last year, it's been phenomenal growth. You know, our customer base has tripled, our installed base has just about tripled. We've shipped almost a half of exabyte of capacity. Yeah, you know, the part on our... Half an exabyte. Yeah. I can't, I'm trying to think of the zeros here. Yeah, come on, Mark, in five years we're gonna be coming back. Ah, the exabyte's the new petabyte, you know. We figured out, we figured out for one of our briefings the other day that we've served, and it's constantly changing, 208 Terra IOPS, which is 10 to the 14th power. You're breaking my brain. So anyway, the short answer is it's going exceptionally well. We're very pleased with the momentum of the business. We're establishing a very strong brand profile, and we're investing where it belongs. And one of the other things that we talk about a lot with customers and prospects who are being bombarded by the all flash messaging, if you look at all the all flash players and you don't even have to name names, their business models are highly leveraged, let's put it that way. And that's a challenge, right? That's not necessarily a sustainable model. We're not in that boat. We are very fiscally prudent. We are very conservative. We invest where we need to invest, and the results kind of speak for themselves. So we're not putting ourselves in a position where we're going to be in trouble down the road from a cash flow or an operational perspective. So, all right. Well, Randy, Infinidat is a relatively young company, but definitely very mature roots in the storage industry. Both of you always appreciate coming on the program and looking forward to we're digging in one of your customers in the next segment. So stay tuned to keep watching the CUBE coverage from VMworld 2016. Thanks for watching. Thank you.