 Live from San Francisco, extracting the signal from the noise. It's theCUBE covering Nimble Storage, the power of predictive analytics. Now your host, Jeff Frick and Stu Miniman. Hey, welcome back everybody. Jeff Frick here with theCUBE. We are live in downtown San Francisco at the Nimble Storage predictive flash launch. Pretty exciting event and we're really excited to have the guy that's running the whole show here, the CEO of Nimble Shoresh, Vasudevan. Thank you, Jeff. Welcome. Thank you very much. Pleasure to be here. Great event. Yeah, I loved it. Thank you. We have people flying in from all points of the compass. We did, I have to say we had press from international locations coming in, we had customers coming in and it's sort of a rolling thunder. We had several, over 1,500, 1,600 people on the live stream. We have events rolling out to many cities. It's a fairly major milestone for us so we're very thrilled about how it's gone. So for the people that didn't get the memo and are just tuning in, what is this all about? You know, I think first and foremost, what we've done with today's announcement is introduced what we call the industry's first predictive all-flash array. It's, as an all-flash array, we really believe we've leapfrogged every sort of all-flash array in the marketplace on scalability of the all-flash array, on lowering the cost of deploying all-flash systems by a factor of 33 to 66% on having a system that's highly resilient, almost six nines of availability. So as an all-flash array, it leapfrogs other all-flash arrays in the market. But for us, and I fundamentally believe this, increasingly to a customer, storage is like plumbing. They care about the application experience. We get caught up in the speeds and feeds of storage but to a customer, it's about how are my applications running and that's really where the predictive analytics part of our value proposition comes into play. InfoSight is our cloud-based predictive analytics software and it's game-changing in terms of predicting what might go wrong in our customer's environment, preventing those issues, 92% of our support issues are ones where we identify what might go wrong and so that's really the two components of our platform, predictive analytics to make sure infrastructure runs smoothly. Flash, whether it's adaptive flash or all-flash managed as one single architecture so you can accelerate every application in the enterprise and that's really sort of in a nutshell what we are announcing. That's great, go ahead. Suresh, you know, you said on stage that one of the key tenants of Nimble is to help reduce the cost of storage. And you help unpack for us because you look at it and you say, everybody uses the same components. Everybody uses Intel, everybody uses NAND from a couple of suppliers and I look at your guys' margins. I mean, you guys are 66, 68% margin, so why can Nimble offer a price point that's better than companies that have a much greater volume than you do? No, absolutely. And Stu, in answering that question, I'll start by saying we've had this kind of a cost advantage on our adaptive flash arrays for the last five and a half years in a nutshell when somebody deploys a hybrid flash array to meet a given performance and capacity and then whether it's EMC, NetApp or others, and they were to configure a Nimble system. Routinely, we used about a fifth of the disk resources, a fifth of the flash resources to accomplish the same, right? And so that's the advantage we haven't had on adaptive. What I'm excited about is that we have a similar, very compelling advantage on all flash arrays and there are four contributing elements that make our all flash arrays much more cost effective and I'm not referring to things like management costs and so on, this is capex of the all flash array. So the first one is data reduction. It's, we've implemented variable block inline data application, variable block compression, pattern elimination, zero copy cloning. So the combination of best in class data reduction, I would argue that our companies like Pure have done something very similar and they would argue are best in class compared to others that have all flash arrays, it gives us an advantage. So that's the first element. The next three elements are very distinct against almost everybody, so let me walk through those. The second element where we really leapfrogged is in the amount of DRAM that we need in our systems to manage a given amount of flash capacity. So every storage system, particularly all flash system, requires a fairly large amount of memory to accomplish two things. One, it needs that much memory to accomplish deduplication because deduplication is very metadata intensive. Second, more intelligent systems often manage endurance in software and to manage endurance and do ware leveling, you need to have a large amount of metadata to know exactly what data to move to evenly distribute data on flash. That, again, is metadata intensive. What we've been able to do is design a system that's very optimized for the metadata footprint. We need about one-tenth to one-thirtieth of the amount of DRAM to address a given amount of flash compared to others in our industry. So it makes our controllers less expensive, and that's the second advantage. The third advantage is we've optimized all along for the use of cost-effective, dense 3D NAND, and so we are able to use extremely low-cost flash, and the way we manage endurance allows us to have a very long life out of that low-cost flash without having to set aside a large amount of overhead. So other companies that do software endurance using 3D NAND also manage to get to use 3D NAND, but they have to set aside a large amount of extra flash to manage endurance. We do that without wasting a lot of flash. So that's the third element. Low overhead, low-cost flash is the third big benefit. You know, one of the most fundamental advantages on the fourth element is we have a single operating environment across our adaptive flash arrays and our all-flash arrays, and a large, I think most customers that deploy all-flash for production would still want to optimize the cost of backup and DR by using low-cost adaptive flash arrays, and that's sort of another element that gives us a cost-effective run. So let me ask, so is that the same operating system, same version that runs between the hybrid version and the all-flash? It is, so there are unique elements, in all-flash arrays we don't use flash as a cache, it's persistent storage, we use some advanced endurance management algorithms, but all the data services, so it's a common operating system that recognizes when it's running on an all-flash array or an adaptive flash array, all of the data services, encryption, compression, replication, snapshotting, cloning, thin provisioning, are all common, right? And so it's a common operating system that recognizes certain unique attributes of all-flash versus adaptive flash. All right, so what about the customer segments? Nimble started with more kind of the mid-size enterprises, it pushed a lot into the larger enterprises, and you talked, it's the number I heard hundreds of service providers today using your products, so how do you manage a solution? Can one product line really cover that broad spectrum? It's a great question, I think, so it is indeed true, we have seven and a half thousand customers, I would say 6,000 are what I describe as mid-size enterprises, still sizable organizations, anywhere between 1,000 to 2,000 employees, hundreds of what we describe as global 5,000 organizations, so large enterprises that have a global footprint, and hundreds of cloud service providers, including SaaS companies and infrastructure as a service companies. And one of the unique advantages we have is it's a broad technology platform that appeals to the entire, to all of those customers, and we've been able to, without making the operating system complex, without making management complex, deliver functionality that our service providers find extremely relevant, as much as large enterprises or mid-size enterprises. Now, there is a, I have seen other companies in our industry try to branch out too broadly with a single platform, and get to a point where it becomes extremely complex and they lose simplicity in pursuit of that, so that's something that we are very aware of, when we feel like, I will say for us, the one thing that trumps everything, that's why InfoSight is so central, is ultimately it's about storage is plumbing, applications are what we care about, simplicity is what our customers care about, not having to manage storage, and so that's where we draw the line between trading off simplicity for more functionality, and we manage that very carefully. So, there are some leading up to this that have been asking, when are you going to have an all-flash? When are you going to have an all-flash? Back in 2014, I asked you, when are you going to have an all-flash? So, what do you say to the people as to, you know, I don't want to say what took so long, but why is it right to move now, and can you give us a little bit of insight as to what you guys were doing behind the scenes to make sure that you did it right? So, we started the all-flash program, I was just commenting to somebody exactly about two years and two months, so 26 months ago is when we started down the path of building an all-flash array, and I will say it was solely tempting to build an all-flash array that was standalone. In fact, the biggest debate we had early on was, do we just build an all-flash array that's a standalone, or do we need to think about common data services with the unified flash fabric? And very early on, we concluded that there was an enormous amount of value to having a single platform with application mobility, single platform where you can replicate back and forth, info site with all the sensors we had built in, we absolutely believe we wanted to leverage. So to some degree, it made the process of bringing the all-flash array to market longer, but we also believed it was the right architecture with which to approach the problem. So that's sort of how we thought about the right thing to build versus getting to market sooner. I'll also say, the good thing or bad thing about storage, there's a reason why EMC, who was the gorilla in our industry for so long, in 1990, 1999, their market share was 33%. In 2015, their market share was 34, 35%, right? So it's storage, unlike networking where customers often say, I'm in the Cisco shop, or they standardize on one. In storage, if you have a better product and a better architecture, it's always possible to go to a customer and say, we are the right answer to the next application you deploy. So to some degree, we don't feel like being late is as penalizing as having the right, as relatively important as having the right architecture. So that's how we thought about what we want to do. Yeah, I actually, I saw an interview that you did and you said over the last two years, there's been more shift in market share than there was in the last 10 years. Then in the last, indeed true. Indeed true. So what is leading to these just tectonic shifts in storage and how do you position yourself there? So it's interesting, I think it is absolutely true. When we describe sort of what God has started, we said there were two catalysts and I have to say my thinking is starting to shift on this as well. So I think one of those catalysts unquestionably is flash and I do believe that architecture matters that if you've designed a system that's optimized for hard to use flash, then the economics of that system are going to be dramatically better than the economics of someone who's retrofitting for flash. So that's one of the challenges that companies like NetApp and EMC and others are facing where they are seeing declining market share. One factor is that they're not as efficient with flash architectures as they could be. Now the second thing though that I think even in the end it's possible that using margins and so on and so forth that they can try and retain accounts but I think the place where it's hard for them to really go is in simplifying their architecture. So one part of it is performance for a given cost and that maybe you can try and work around. The part that's much harder to overcome is that when I manage an EMC environment, when I manage an NetApp environment, I need train storage admins. When something breaks down, it sometimes takes hours to days to root cause what's going wrong. I spend gobs of time gathering logs and dyags and waiting to speak to an expert that can tell me what's going wrong. That complexity I think is frankly what's even more daunting and why I do not believe it's going to be easy for these companies to regain market share. And that's why I think the shift is happening so radically and so quickly. So when we look out of the marketplace, there's two trends I'm wondering if you can comment on. Public cloud and hyperconverge. Yeah, both great questions. So let me take the public cloud questions too. Look, I think the public cloud is a force that is here to stay in a really, really big way. And when I step back though from the point of view of a customer and I think about deployment options for infrastructure, I basically see any enterprise managing four deployment options. The first option is SAS, probably the best and fastest way to minimize complexity of deploying applications. And that's sort of rapid adoption where it's possible for someone to deploy SAS, they're going to deploy SAS. The second, the next three are infrastructure as a service. So where I'm still following classic data center based or standard space way of deploying infrastructure, but instead of managing it on my premise, I'm going to a terror mark or a century link or someone else. The third option is on-premise deployment. So same as in the past. And the fourth is public cloud. Now, I believe that public cloud will grow, but it's hard to imagine sort of public cloud going beyond a certain percentage of how a customer approaches deployment. And here's I think where public clouds will tend to remain somewhat of a challenge for many customers. In order to truly leverage public cloud, you have to have application development resources within your organization, because often the ways in which you're building the apps need to adapt to public cloud deployments. And that tends to limit how far you can go. That's one limiter. The second limiter is often it's economical when you're thinking about content depots and object storage based data models. When you're thinking about transactional applications, adapting those applications with public cloud tends to be a bit more daunting. And economically it's not entirely clear what the benefits are, compared to for example managed private clouds. So that's how I think about the impact of public clouds. It's clear that some part of the infrastructure will go to public cloud and that part will not come back to storage systems vendors, engineer systems vendors. But when a customer goes to SaaS companies or infrastructure as a service companies, those are companies that are still deploying products from companies like us. And so that's how I think about the cloud opportunity. Just to sort of synthesize for us, deployment into SaaS companies and infrastructure as a service companies is one of the fastest growing segments. As I mentioned, we have hundreds of service providers. It's already in the high double digits as a percentage of our business is in selling to these service providers. So that's a fast growing part of our business today. Great, all right. So then the second part, hyperconvergence. So to me it's actually interesting. Hyperconvergence is, we often talk about the motivation for hyperconvergence as Google like infrastructure where essentially they set for problems like map reduce, deploying server combined with storage. That was the inspiration behind Nutanix's hyperconvergence for example. And the benefit that hyperconvergence brings is extreme simplicity where I don't even have to worry about managing compute decoupled from storage. Now the challenge that hyperconvergence brings with that simplicity is that by pooling compute, storage capacity, storage IO bandwidth, all into one single building block, scaling becomes expensive. Whether I need to scale compute or capacity or IO or network bandwidth, I'm having to add more nodes at a time. So the trade-off tends to be economics of scaling on the one hand, versus simplicity of management on the other hand. And that's really how I think about sort of the pros and cons of hyperconvergence. So given that there are certain environments where I believe hyperconvergence will start, will do well when your application is homogeneous. So VDI is a good example where every time I need to scale compute I also need to scale storage. It tends to move in lock step and that's where hyperconvergence does do well. Other environments, customers still want the benefit of converged management but don't want to necessarily trade-off efficiency and that's where I think best of breed converged systems will do well. Things like FlexSpot, VBlock, we've partnered with Cisco to do SmartStack. Those approaches where you're trying to bring the simplicity of converged management but retaining the ability to scale independently will continue to do well. Just as I said, I'll point out one thing. The motivation for hyperconvergence was Google-like architectures. Increasingly, open compute is moving to a model where you're essentially decoupling compute from even memory and storage. They are also moving towards a model where they're saying combining the hardware is not what we need to do. Converging management but scaling hardware building blocks independently is the right long-term architecture. So to some degree, converged infrastructure of the kind that we're doing with SmartStack with simplified management is where they are also starting to head. So first I want to shift gears a little bit and talk about your CEO hat, running the business. A lot of conversations in the Wikibon community about kind of new age storage, new technology, new innovation, but it's hard to get to escape velocity. And that's the exact term we use quite often. You made it, right? You IPOed, companies often running, 7,800 customers, how did you do it? What makes you guys special? How are you kind of getting by in this not-so-friendly market as it was from the year six months ago? So a lot of thoughts there, Jeff. So I'll say the first comment is on sort of what it takes to get to the point of being a public company. What is necessary is you have to be sort of sound on certain core aspects, fundamentals of the business model. The ability to acquire customers quickly. Over time, the ability to demonstrate that you're not just driving top-line growth but generating cash and operating leverage. And so all of the having a healthy business model, a differentiated product, and proof points and winning customers are necessary. Unfortunately, there's a reality to market timing as well. Just having all of those attributes doesn't necessarily mean, I know the truism is great companies can go out at any time, but it takes a lot of courage to go when the markets are not in a friendly place. And there are some good companies in our industry that ultimately I'm confident will go out. Timing has to work out as well, right? So that's my first observation. Now the challenge with storage is that the number of companies that have all the right preconditions and the timing will come sooner or later can be numbered in less than one hand, right? So there are a few companies that have that kind of critical mass where when the timing is right, they have enough escape velocity to make it as a public company. For the dozens and dozens of companies, unfortunately I actually think there was so much venture capital available over the last five years that growth was almost funded without necessarily all the preconditions being in place. And over the last six months, that has dried up. I'm new of many, many VCs that would consciously say, stretch your funding now in the current context rather than sort of raise more money to drive growth. That makes it very challenging because I don't believe, I fundamentally believe in a couple of years you're going to see a handful of systems companies continue to play aggressively in storage, a handful of next generation storage companies and dozens of the other early stage companies will find it really hard to get out on the other side of the next two years. I really believe that. So Suresh, I guess with the time we have left, why don't you talk about kind of the company going forward? So, you know, storage company, storage is changing a lot, you've now got all flash array, but InfoSight I feel has given us a little bit of visibility as to going beyond some of just the storage stuff. What do we expect to see from Nimble going forward? There's some companies that are, you know, trying to go against virtualization, some that are building up the full stack. You know, what will Nimble be when it moves to the next phase? No, it's a great question, Suresh. Our dream is simple. We start with a belief that fundamentally a new generation of storage companies will evolve as the leaders over the next two to three years and that the traditional leaders will find it hard to maintain their position over the next three or four years. Now the real question then becomes who can claim that mantle of next generation leader? And there are some preconditions that I believe are absolutely necessary to being that next generation leader. One is breadth of technology. If you're going to say I want to display someone like an EMC at a large enterprise, addressing all of the workloads that an EMC has been able to address requires an extremely broad platform. And it's not a daunting task. And by no means done, there are a lot more, many more innovations that I think we need to work on in order to be able to claim that we can displace everything that an EMC can do. So continuing innovation to address different forms of deploying storage, different forms, different workloads where we can take on every workload, that's one step. The second, we are proud of having a quiet seven and a half thousand customers to claim critical mass. You have to be in the tens of thousands of customers. So customer acquisition at an aggressive pace is a second dimension that we have to really, really go work on. And the third ones, Stu, I would say I want to measure success three, four years from now on what kind of a market share and what kind of a business have we built. But the other measure of success is really turning around to the organization that we've built and having every employee at Nimble say, you know, I've worked at a lot of companies, but one of the best places, the my best experiences was at Nimble storage. Maintaining the culture of the organization as we grow in size is a third dimension of success. All of this in the context of the storage industry. InfoSight, I don't want to say a whole lot more Stu, but I will say managing storage and infrastructure from the vantage point of storage is just the beginning for InfoSight. I fundamentally believe that as you look ahead, we talk about IoT as a major trend. Every consumer device is going to send information back to the manufacturer. I believe in the data center, there will not be a single device that is not proactively reporting back on its health and the health of everything surrounding it to the cloud and someone else is going to come in. Customers will not have to call in to ask for help anymore. Wenders should be calling customers to optimize their environment. And so we are just scratching the surface of where we think we can go with InfoSight. Not just in storage, but beyond storage. I hope you guys can lead the way for just getting rid of level one and level two so we can go straight to the experts when we need something. Absolutely, no, that's our goal, that's our goal. All right, Trash, well thanks for taking a few minutes. Congratulations on the great event. My pleasure, thank you so much. Absolutely, and you're already there. You're already telling your customers 92% of the time before they call you. That's right, no, thank you very much. It's a pleasure. Awesome, all right, Jeff Frick here with Stu Miniman. We're at the Nimble Storage Predictive Flash launching downtown San Francisco. You're watching theCUBE. We'll be back with our next guest after this short break. Thanks for watching. Since the dawn of big data.