 To EMC World, this is theCUBE, our flagship program. We go out to the events, extract the signal from the noise. I'm John Furrier, the founder of SiliconANGLE. I'm joined by my co-host. I'm Dave Vellante of wikibond.org, and we're going to talk Flash. Zahid Hussein is here. He's the Senior Vice President of EMC's Flash Division. Zahid, welcome to theCUBE. Thank you, good to be here. Great to see you again. We first met down at the Financial Analyst Conference, and then I was out in your offices, we did a little deeper dive, really impressed with what you guys are doing. John and I were walking the floor last night, we saw your booth, you know, we woke up this morning on the door of our hotel room where these extreme flash branding things, you guys are going for it here. A lot of customers, tell us what's going on at the show, tell us how you feel. A lot of exciting stuff happening. We feel great about the portfolio that we're bringing to market. We're seeing that there's a transformation occurring in the data centers as you move from to a flash-based architecture and it's impacting applications, it's impacting operating systems, it's impacting networking infrastructure, and of course there's a ground up rethinking of storage architectures that's occurring. So it's really about bringing business value, business advantage through performance, but also addressing total cost of ownership. And what we want to help our customers and partners understand is that we're there to work with them to help them transform their data centers. We asked Pat Gelsinger when he was at EMC at the time, when he was president of the product group a couple of years ago, that's our fourth EMC world, you know, about Fusion IO. And Fusion IO was just about to go public and he said, you know, oh, we're watching it the next year, they went public. Pat, you know, what are you going to do? He said, we are going to be in there, we're going to be, we're going to, we're going to essentially have flash. So since then you guys have made the big moves. Yep. But I really want to ask you is how do you see this really being a catalyst to accelerate the transformation? Because we've covered Fusion, Violin, all the big players, you guys have been covering you guys like a blanket. Obviously flash is changing the architecture. So there's a conversation around architecture fidelity. Yep. Okay. And also building out for this modern era for the next 10, 20 years, it's been this inflexible point. How do you guys look at that? Obviously it's a mindset change. We just had Vick on your CIO of EMC, he's talking about a whole other mindset, he's coming from GE. This new mindset's critical. So how does flash, one, accelerate change? And two, what mindset shifts will it cause that you guys want to either accelerate or promote? So let's go back to the first part of your question around the mindset shifts and then what gets accelerated. But mindset shifts are, one, the performance you're getting out of flash isn't just about, hey, here's a faster version of your storage, it's going to change the way in which you're analyzing the data, operating on that data, the number of simulations you're running in parallel, your response time expectations, everything that you expect is going to change. And it'll change by an order of magnitude. When things change by an order of magnitude and expectations, they start to change behaviors. And then applications themselves come to expect something different from the infrastructure. And so you'll see applications themselves start to morph with a different expectation of what the response times or the capabilities are of the infrastructure, of what the storage is. So it's a very virtuous cycle. You'll see the infrastructure change, applications change, as applications change, it accelerates the changes in the infrastructure. We see that accelerating over the coming few years. And that's, I can't predict exactly how fast it'll happen, but I think that what we should expect is that over the coming few years you'll see exponential changes in deployment of flash at various parts of the infrastructure in the server, in arrays, and that'll have deep implications, profound implications on the rest of the networking infrastructure, the computer infrastructure, and in particular applications in the way that users analyze their data and what the response times and the number of operations they perform on data. So everyone likes to get on the speeds and feeds. And Dave and I have talked about this in the queue all the time. We've moved from ports and speeds and feeds to business conversations, but you can't help but talk about some of the economics and the performance improvements with flash. So the question I want to ask you is, obviously IOPS and latency are kicked around. So that's cool. That's today. That sets the stage. What other metrics, well first of all, latency or IOPS, which ones do you think is more important? And then two, what other metrics post speeds and feeds do you think about when you look at what flash is going to be doing and evaluating what are the key metrics around beyond IOPS and latency? So between IOPS and latency, we tend to look at latency as the primary driver. I think that as you drive your latencies further and further down, you'll get best utilization out of your ever better packed multi-core CPUs and the utilization will continue to increase. So the latency, driving down that latency allows you to not just use flash as a storage tier, but it also becomes interesting as a memory tier. So we're going to pursue that. IOPS of course are interesting, but it's a way of driving more consolidation. IOPS is something you can scale. You can't scale your way to lower latency. So we're going to keep on driving to lower and lower latency. And that has the great side benefit of getting to higher IOPS. The, now I think the- And latency points to more of the kind of what's relevant around mobile apps and kind of what's happening in the real world. Absolutely, you know, you're getting this, not just a great amount of human generated data, but you're getting a lot of event data coming in and machine generated data. And you want real time analysis of that data. Only solid state storage will give you the ability to go ahead and capture and analyze that data at scale. In terms of where the industry is going and what is the kind of rate of adoption of these things and where the price point's going. We are seeing that just Moore's law as you go from what was single level cell flash a few years ago in the enterprise to multi level cell now, which is a cheaper flash to really going to consumer level flash and becoming really now and going forward. We should expect that flash pricing will continue to drop, that a large class of workloads will fit very comfortably into all flash environments. And you'll start to see customers that are really going for many petabytes of flash in their environments. And we're starting to see that now. We're getting interest from large enterprise customers in going to petabytes of flash storage. So you were sort of using this metaphor of a wheel before I was imagining a flywheel that could spin faster and faster and faster as things progress. So you mentioned economics, certainly application performance and I wonder if ecosystem is in there. I'm thinking about things that, what are the things that can make that flywheel spin faster? One of the things that you clearly get from flash from solid state storage is a reduction in power. So the total cost of ownership starts to become interesting. So what starts to spin that flywheel is not just that it's faster, but your power is reduced, the amount of time that you're spending configuring, tuning your environment comes down so your operational expenses come way down. You start to see that the entire total cost of ownership of deployment of a flash-based environment starts to become very, very compelling. And so that's not to say that all of your data is going to reside on flash because the amount of data being generated is staggering. And there's no way you're going to put all of that data into a flash-based environment, but I think we will start to see that the cost effectiveness of putting your hot data on flash is going to become extremely compelling. I think it already is, but it'll become yet more so. As you look at the full suite of capabilities, what's the best way to utilize your servers? How do you get the most out of them? What is the power, what are the power requirements? And then what are the operational expenses of going ahead and configuring, tuning, doing all the performance monitoring of your environment? Flash gives you predictable consistent response times on that. So we all know that applications are constrained by mechanical disks. Your company has to do some unnatural things to make mechanical disks run faster, things like short-stroke. And we'll look back 20 years and we'll look at these crazy things we were doing. But nonetheless, EMC makes a lot of money doing those unnatural acts. Are you under pressure not to cannibalize those existing products? Can you talk about that a little bit? I think that we are under pressure to innovate and where it makes best sense for our customers to do what's right for the market. We're not under pressure to go protect our existing products. We're under pressure to do the right thing for our customers. And when it makes sense to do the right thing for our customers by bringing them flash-based solutions, we will do that and we will continue to innovate in that space. There is, very candidly, there is overlap between some of our products. But I think that we are much better off having some degree of overlap than having gaps on our product line so that we can bring really full solutions to our customers. That's really what we're focused on more than anything else is, what is the right thing for our customers? Well, EMC kind of has the heritage of doing that. You have a lot of internal competition and I think that's healthy. John, what's your take on all of this? You know, I always love the flash and the whole solid-state innovation because the whole thing, Dave, even three years ago when we were covering Fusion, the Naysayers were like, oh, it's too expensive but we were riffing on the whole scale-out open-source thing saying, hey, one sand can really explode in value. And so what that enables is a new mindset and a new business model, quite frankly, in IT. So to me, I think one, this is going to increase the IT IQ in terms of forcing IT and anyone doing any kind of infrastructure to be smarter about how to build architecture and for future architecture. But Dave, for me, the parallels big data. And when we were at Hadoop World four years ago, the first, you know, Hadoop Summit in Hadoop World, we talked about big data and then last year, you made the point about big data creativity. So to me, you know, the quote that we had from, I think it was from someone at Hortonworks said, the only thing holding us back is our ability to be creative and understand what hasn't been built yet. And to me, what Flash truly is, Dave, is that this, what big data is allowed for new things to be, new questions to be answered and new creativity is ultimately the key. So I want to ask, and I think Flash is absolutely in the same track of, oh my God, I couldn't do that before. I'm going to actually replace the memory tier. I'm going to actually increase my memory tier because I'm going to write my software differently. I'm going to build my mobile apps using Node.js and I'm going to use AWS and I'm going to roll a private cloud together and I'm going to blow out the CRM and install something new. I mean, I'm making that up, but you know, that's the kind of the mindset. So the question is, what do you guys see as the new, the new kind of future? I know it's still elusive right now, but I mean, beyond IAPS and latency, assuming that software is going to be written differently with an expanded memory tier, assuming that open source continues to accelerate, assuming people want trust in their cloud, as Jeremy Burton would say, what are the new creative opportunities? I mean, you guys got to look at the horizon. It's a, you know. I think there is, we spent a lot of time thinking about this and doing work on it. There is a class of applications, a new generation of applications that increasingly operates off of a memory tier and runs a set of data services at the application level. And what we would expect, that many of those are being driven by open source community, hyperscale environments, and they're starting to move into the enterprise in various ways. What we want to make sure of is that from a resource management perspective, what is the best way to go ahead and access that, have abstraction models into that, get the lowest latency, be able to scale that, provide levels of protected persistence where it makes sense, and be able to tier that to capacity pools, we provide tremendous value on how we do that. And a lot of our value is going to be in constructing those software layers and the software layers that allow you to scale it, protect it, provide the abstractions, and then be able to really connect that to those capacity pools. Wherever those capacity pools may be, and they may be capacity pools that are local, or they may be remote cloud-based capacity pools. So we will innovate tremendously in all of those things. I smile because my CS degree was back in the days when operating systems designed systems, and Paul Moritz announced in 2010, when he was at VMware, the cloud operating system basically is what we were saying. So this is what you're talking about is looking at the holistic picture and saying, it's now an operating environment with subsystems, software layers. Is that what you're saying? The server node is, I would say that we've kind of evolved past the point where the server node is the key operating system and the data center itself is what you want to look at as the operating system that you're managing. So you've got these very large scale, sometimes shared nothing infrastructures, sometimes shared a lot of things infrastructures, but how do you really resource management across, resource manage across all of those and provide the various tiers of performance and data access at the right abstraction levels? And how does that then move into new APIs, SDKs, scheduling algorithms at various levels, OS or application level, and that's where we want to participate. That's where we want to innovate. And so with that, obviously software, we just talked about the memory kind of tiering, but beyond software-led infrastructure, as we call it, Wikibon, what are the software things that need to get done? I mean, obviously, we agree with you 100%, by the way, that data center is the operating environment and there is a subsystem elements and all that stuff. But what needs to get driven? I mean, software is the key. Software-defined storage is the big buzzword here. Obviously, it's not, you know, software-defined networking with Nasirah. What needs to get built? What is the key enabler on the software side that Flash will really put Flash, you know, in the Hall of Fame of IT? There are, well, but, you know, there are a lot of things. So it's hard to pinpoint it to one, but I would say that of the few that I would name right now, getting to a model where you can continue to derive lowest latency in a scaled environment is going to become very, very important. Driving to a model where you have good affinities for your workloads and that tier of very low latency persistence, that low latency memory tier or storage tier is going to become very, very important. So how do you manage the affinity and how do you manage the scale and then how do you manage reliability across all that? Flash is still a very difficult medium to work with and it just gets harder as the geometry shrink. It is hard and we are investing a lot in making sure that we understand both the low level characteristics of it as well as deal with the higher level abstraction models of how do you scale it and how do you have applications make best use of it for low latency access, for memory based access as well as through the storage stack and how we handle that through the storage stack. Well, we have our all flash arrays, ExtremeIO and then there's sort of things that we're going to do that are going to give you memory tier abstractions as well, but the innovation on the software I think is going to be the affinities, the scalability and then getting to a resource managed abstraction model and a mobility model that makes a lot of sense here. So that's hard. Just a follow up on that. So you're talking about the memory tier, you're talking about doing things like bypassing storage protocols and atomic rights and things of that nature, correct? Am I getting that right? Yeah. Okay, and that through, for example, industry standards like NVM or other standards that you're going to try to put forth or both? So we are working with standards on NVME but I will say that we will do what we need to do to push the technology forward and as much as it's important for us to be driving the standards, we know that sometimes driving the standards means coming out first and then getting the standards to be adopted but really we want them to be standards. Okay, and then my other question is where do you guys fit in the whole software defined storage play? Talk about that a little bit. EMC is, we'll be talking a lot about our software defined storage strategy at the CMC world and you'll hear a lot more about that and things that we're doing around both the controller side of things as we're able to manage a sort of heterogeneous array types in terms of provisioning and then eventually data services in a common way. There is a software defined storage that often means what's happening in a currently software provision model on often commercially available off the shelf servers and that's also a lot of interest to us and many of the things we're doing on the server side with Flash, we see intersections with the software defined storage as well as where we go with software, with server based Flash and the hard part is doing that in a way where you're still taking full advantage of the latency characteristics of Flash so what you don't want to do is build a software provision model and then lose all the advantages of your water and water it down and you can do that if you don't really have a good understanding of what the protocols are, what the networks are, what the affinity models need to be, how to best place your workloads, all those things and that's the trick to it. So from a stack data management storage management volume management stack standpoint, the good thing is you guys are new technology so it's fresh, it's modern, the bad thing is you haven't been around for 10 years developing all those capabilities but then the flip side of that is good news is EMC has a lot of that stuff so can you talk about your storage management, data management stack and how that will evolve? The storage management stack, you'll hear about a lot at the CMC world around the things that we're doing to continue to make it easier to provision, make it a much more of a self-service model when it makes sense to drive a simplicity of provisioning across different array types with different characteristics, make that as common as possible in terms of a common set of APIs. You'll also hear more about how we're making it possible to migrate data across the different array types and we're not constraining ourselves really entirely just EMC arrays, we're opening that up to even non-EMC arrays. The characteristics of flash-based environments make it such that the provisioning and the policy-based models and the capabilities are something that we're going to have to bubble up, make sure that when you're provisioning you have a good understanding of what the capabilities of the flash architectures are and provision according to those. So that's another piece that we're really working on is presenting those capabilities, matching that top-down to the policy and making sure that there's automation of capabilities to policy, driving that to how data is provisioned and how it's moved around the data set. Thanks for coming on theCUBE, we got a break but really amazing content. I'll see it's cutting edge on the front end and really the explosiveness of this mindset of an operating system where the systems guys are back in favor. The big data guys are all about, if you're a database guy, you're hot, big data's good for you, now if you're a systems guy, a lot of engineering to be done but the future is very bright and it's early stages, you guys are laying down the groundwork at EMC for a complete re-architecture, congratulations and again, the future is unwritten at this point and you guys are setting the table, so congratulations. Thank you. This is Silicon Angles theCUBE flagship program, we go out to the events extracted, see them from the noise, I'm John Furrier with Dave Vellante, we'll be right back with our next guest after this short break and here at EMC world.