 From the SiliconANGLE Media office in Boston, Massachusetts, it's theCUBE. Now, here are your hosts, Dave Vellante and Stu Miniman. For several years now, the analysts at Wikibon have been talking about taking the cloud, the public cloud operating model and bringing it to your data wherever that data lives. Hi everybody, this is Dave Vellante and I'm here with my co-host Stu Miniman. Welcome to HCI, a foundation for IT transformation and we're here with Chad Dunn, who's the Vice President of Product Management and Marketing at Dell EMC. Chad, good to see you again, thanks for coming on. Yeah, glad to be here, good to spend time with you guys. So we talk a lot about, you know, VxRail, speaking of foundations, give us a quick update. You know, what is it and what's new with VxRail? Okay, well big news in VxRail land, right? We just completed our transition onto the 14th generation of Dell PowerEdge servers. So this gives us a substantially more powerful platform, substantially more predictable performance and a lot more configuration options that make it fit a lot of different workloads that our customers have. So it really makes it prime time for HCI. So where's the power and performance come from? Is that predominantly kind of new compute? That's a big piece of it. Some of that is software as well, right? VSAN underlies VxRail as a software defined storage layer and we've seen pretty amazing increases in performance just from software from our 13G to our 14G transition. But when we look at that performance now on 14G servers with the Intel Skylake chipset, we're seeing two X performance over the last generation and we're seeing latencies that are very, very low. And that has to do with more and faster memory channels, more threads, overall faster processors. So really off the hook in terms of performance that we're seeing. Chad, when we look at HCI, it's really about the software layer. Often it gets overlooked. What actually has to happen between the software and that underlying hardware? Are there optimizations? Does it matter if I'm using the software? What's optimized for that next generation Intel chip? Yeah, it's all about the software. Or so software vendor would say, but we know that when you're treating something as a system, you need that hardware and that software to work together in perfect unison as a system. And we've done a lot in this generation working with the PowerEdge team to make sure that we have the right hardware hooks and design points that are focused on HCI. That goes from things like the devices that we used to boot up and where we would execute the hypervisor kernel to network connectivity. And really importantly to the in-band channels that we used to update all the little pieces of firmware that operate the hardware inside the system. You need to be able to treat those as a system, update, lifecycle manage those all in context of one another. So having direct and deep meaningful access into that hardware is critically important when you're operating a system like this. Yeah, when we look at kind of our cloud strategy in general, it's about the data. And when we talk about data, it's things like predictability and latency. It's about kind of the power of the underlying thing. Maybe give us a little bit more specifics as to what you're getting in this generation. So the big difference here above and beyond the performance which is about 2x what we saw from the last generation if we look at the same hardware or the same software running on the two different pieces of hardware, about 100% better. That's really just part of the story. It's the predictability of latency that's critically important. If you're gonna migrate tier one workloads under this infrastructure, you need to ensure that other workloads are not gonna disturb that performance. So when we look at this, we look at how the IOS per second increases and we look at the overall latency. How long does that latency line stay flat? So when we look at this generation, we see over 2x the IOPS, but the horizontal line where we look at the response time and latency, it stays flat nine times longer in this generation than in the last. So you've got that sub millisecond response time, even a very, very high IOPS. You can put a lot of different workloads on that same infrastructure, still get predictable performance. I think the other thing that people understand is that HCI it's just like, it's that little Lego block you build. But it's not just one Lego block. What do you see from customers? What's kind of the portfolio? What are the decisions that they have to make to kind of pick the right configuration? So when you're a kid and you get your first Lego set, you get a lot of pretty generalized blocks. They're all square and some are rectangle, but not a lot of variability. When you get up into the big leagues of the Lego Star Wars set, you've got a lot of specialized parts and you can do really advanced, really cool things. That's really where we're at with HCI right now. If you wanna really tune the infrastructure for the workloads that you have, you need a lot of variability in the processors that you choose, the amount of memory, the speed of memory, and even the storage. It could be hybrid. Some people still choose hybrid HDDs, but even within Flash, people will choose SAS or SATA drives depending on the performance and cost benefits that they wanna realize. So being able to scale up and down the processors, the memory, different types of storage is critically important so you can fit into those different workloads. Also, a lot more people use this for VDI and for high-end imaging. So the ability to pack these things full of graphical processing units and still be able to power and cool the things is critically important. We have a lot of applications in those verticals where there's video processing and these are required. So we don't just have one model of VX-Rail. We've got a number of different VX-Rail models, all of which can scale up and then of course HCI intrinsically can scale out. So that lets you really fine tune and get to that expert level in terms of your Lego building blocks. So Chad, a minute ago you mentioned workloads. So as you're bringing this sort of 14th generation server technology to VX-Rail, how has it affected workloads? What are you seeing as the sweet spot for workloads? So if I were to think back a year, the question that every customer would ask is how do I know which workload is right for HCI? And a lot of times they even lack the vocabulary and taxonomy to say, okay, that fits, that doesn't fit. What's happened in the meantime though the software has gotten so much better, the hardware has gotten so much faster and more predictable that the question is, well, what workloads are not right for HCI yet? And there are very few that aren't. So we've seen people generally start off with one workload. Maybe it's VDI, maybe it's a database. And then they start to move other, as they get comfortable with it, they move other workloads over to it. Obviously we've got a big install block or install base of VX block and V block. We see a lot of those customers start to migrate workloads from there onto a layer of HCI and more and more those are becoming tier one workloads. Crate and Barrel is a great example, a great customer of ours. They're moving their point of sale systems onto VX Rail. Now for a retailer, your point of sale system, that's about as mission critical as you can possibly get. So they and others now have the confidence to start to move these things over. The only outliers that we see are some of these very big data applications that are hugely right intensive and we actually usually end up selling a layer of hyperconverge with our Isilon arrays. They store that data and then put a layer of hyperconverge compute around it. Because in some ways hyperconverge is just a better way to server, if you know what I mean. What if we could talk about the business impact? What are customers seeing? How are they quantifying the value of these systems? Share some stories or color there. Sure, it's all about operational expense savings, right? How much more efficiently am I going to be able to operate this infrastructure? It's not so much about capital acquisition costs. So when you look at the typical operational expense savings and that comes from us doing all the life cycle management of the hardware, of the software, of the cluster as a system, you see those costs go down. Really good example is first credit of British Columbia, another one of our good customers. Now they've deployed this, they've seen 30% OPEC savings and they've seen 50% power and space savings. So you get a smaller package because you don't have separate storage array, separate servers, but you also have really one function that needs to operate your environment and that's the virtual administrator. He or she is the one that really operates everything. You don't have separate storage, separate compute, separate virtualization teams that have to look after the infrastructure. So yeah, first run is very easy, very fast to deploy, but it's day through, day two through day 700 and day 900, where you see that recurring operational expense savings where it really pays off for customers, all the updates in life cycle management. So Chad, you talk about the excess and all the customers, what about the customers that haven't looked at kind of the HCI space yet? What are they missing? What do you say to those customers that maybe, aren't sure if the water's right to jump in yet? So there's really three ways that you're going to encounter a customer who's going to consider HCI. You're either going to refresh a server, your servers are up for maintenance and you're going to take a look at HCI as the next step in your evolution of your compute strategy. Or you're going to refresh your storage and you're going to look at hyperconverges the next step in the evolution of your storage strategy. Or you've got that one workload that's probably net new and it's going to be sort of an isolated case and they need an infrastructure and they need to stand it up fast. That third case is really the one that drove the initial adoption of HCI. I can't tell you how many of our customers started with VDI. I mean it's so cliche now to talk about VDI as a killer app for HCI but that's how so many people started because it's a very bound and isolated infrastructure and from there they get comfortable with it and they start to bring other workloads onto it. So if you're thinking about refreshing your servers, if you're thinking about refreshing storage it's time to kick the tires on HCI. If you've got a workload that you need to stand up quickly and you don't know how big it's going to be one, two, three years down the road HCI is another, it's another opportunity to look at HCI because you can start with a very small infrastructure but you can grow it to a very, very large one. I want to talk a little bit about digital transformation. I mean everybody's talking about digital transformation and to us digital transformation is all about how you leverage data and the edge is exploding. We've envisioned sort of a three tier data mall yet the edge you got maybe an aggregation point and you bring it back to the cloud and that cloud can be a public cloud or it can be on-prem. So you've got to have some kind of cloud infrastructure to manage all this data. So where does this fit in the context of transformations and why does hardware matter? Yeah, well let's go from the end and work back to the beginning, right? Hardware matters because of form factor for one. As you start to push compute out to the edge, right? You want form factors that are small, don't consume a lot of power but still have a lot of processing power and can manipulate that data. The whole internet of things phenomenon that is creating all this data out at the edge presents us with a conundrum, right? The data itself is not that valuable. The insights that we get from the data are immensely valuable. Bringing all that data back to the core to do something with is not cost effective. So it's how do we turn the data at the edge into information and then how do we funnel that valuable information back to the core and leave the unvaluable data out where it is? Hyperconverge fits really well there because you can have devices of very small form factors that are very quick to deploy, very easy to manage remotely. At the aggregation point, you can have simply larger versions of the same thing or more of the same thing and then finally at the core you can have very large clusters of hyperconverged appliances like VxRail to do your processing. Now the key is from an operational perspective you've still got a single pane of glass that manages everything, right? It's still the same set of tools, it's still the same hardware and software lifecycle management process that happens out at the edge at the aggregation point and at the core. So again, it comes back to the operational expensive, making decisions closer to the data and then managing everything with a consistent set of tools. So I wonder if we could also talk about the competition and when Stu and I think about the competition and this sphere we look at, well first of all this is all sort of software defined, everything's moving to software defined. So we see two vectors. One is head to head competition with other software defined suppliers and the second big competitor is hey, I'm just going to roll my own. So let's start with the former. Why Dell EMC versus vendor A, B, C or D? Sure, sure. It really gets down to what your goal is as a customer and we obviously have multiple options within our own portfolio and those perfectly find solutions for a lot of people. But number one, if you're a VMware user and you want to optimize around the VMware user experience, then VxRail is the way to go because we do co-engineer this with VMware. It's not just a regular partnership. We have engineers and marketing people and product managers at VMware that functionally roll up to our team and so we do behave as one engineering, one product management organization to really optimize the user experience for VMware. Secondly, architecturally from a vSAN perspective, this is a service that's baked into the kernel of vSphere. So in terms of performance and the overhead that it creates on CPU, memory, et cetera, this is the best game in town. We can do more IO, more predictably with flatter latency than really any other solution that's on the market in the HCI space. Every other one takes a virtual storage appliance approach where they have something running on top of the hypervisor. The very long and circuitous data path will performance test against solutions like that all day long every day. That doesn't worry us at all. So again, if you're a vSphere customer, VMware customer, it's the most obvious choice and from a performance perspective, you're not giving up anything, right? We don't want users to have to sacrifice the storage functionality or the performance of the compute functionality. Just because a type of converge and you scale out doesn't mean you can compromise on any of those axes. Okay, what about the guys who like to change their own oil in the car and the spark plugs and tune it up? Sure. They want to roll down. It's been a long time since I've been able to work on my own car. But all right, so I encounter these kinds of customers all the time. It's the build your own crowd and it's what they've been doing for a long time and it's great, right? I build my own computers at home, right? And I have my own ESX server at home that I put together. I can't afford it next round. But there's no employee discount. So I'll tell you a story that sort of will hopefully make sense, right? My first job when I got into this business, I went to Boston College. My first job in work study was to keep a spreadsheet that had all the MAC addresses and all the IP addresses for every host on the BC network and keep those in sync. You're really good at that, I bet. I was excellent at that. That is not a skill set that is in demand right now or really even at that time. But when you think about what it means to take a software-defined storage product like VMware vSAN and take an X86 server and put those together, yes, you're getting to the same destination of running vSphere on a host with software-defined storage. You're missing the system-ness, right? We go to a lot of trouble to make sure that we're managing all of those things in context at the cluster level. All the little pieces of firmware, and there are roughly 12 or so pieces of firmware that we have to take care of from the BIOS to the drive controller firmware, the drives, the boss card, which is our boot media, the iDRAC firmware, the backplane, power supplies. At EMC, in legacy EMC, we spent 30 years building arrays. We had all those same challenges with all the different pieces of firmware and software that all had to function as a system, and we did that. And we guaranteed that it would live up to five nines of availability for the customer. That's exactly what we do when we deliver VxRail as hyper-converged. If you want to choose to build those things yourself, that's fine if you have the skills and that's how you want to operate your business. The five nines is now on you, though, right? Because you're the one responsible for bringing all those parts together. So yeah, it's certainly a valid path for others, but the market is shifting and we see more often than not people are moving toward a buy approach rather than build. You bring up a great point. I remember back in the early days before we even called it HCI, you think about VxRail, oh, well, is the storage admin going to buy it? Is the virtualization admin going to take that over? What's excited me about this wave is the, oh, here's the cool stuff that companies are doing now that they're not spending their time keeping spreadsheets. Okay, exactly. What is the owner of this look like in your environments and any cool stories you're hearing from customers transforming their organization? By and large, the operator of this is your virtual admin. The person who is at home in vCenter, in VROPS, with maybe even VRA if they're going full infrastructure as a service, that's really the user of this, right? And the dynamic you mentioned is similar to what we had with Vblock, right? Customers who went Vblock who said, I'm going to change my operating model to a virtual administrator versus compute storage network. Customers who didn't change the operating model were not happy Vblock customers. Ones that did change the model did. And I'll tell you, I'll tell you a real off script anecdote. Recently I was traveling in Europe and I started playing a game with the sales guy we were traveling with because in Europe very often, they have more of an affinity to putting their logos on the sides of buildings in a lot of European cities. So as we would go to these different cities and we went from Stockholm all the way down to Rome, to Switzerland, to Amsterdam. Yeah, we're just spotting VxRail customers. Who's going to spot the most? And the one really interesting one is we checked into a hotel late night in Switzerland. Next morning we meet for breakfast and he goes, did you spot the rail customer? Said who was it? We went into the bathroom and they have these squeeze bottles that have the soap in the shower. And it's a cosmetics company and they're located in Germany and they do obviously a ton of business all over Europe and they had outsourced a lot of their IT because their core competency is not IT, it's cosmetics. And they now have one guy that looks after all of IT for this company rather than outsource it to two different companies to manage all this and he runs it all on VxRail. So transformative, yes, to that company very transformative at a very small scale, but that pattern sort of repeats itself the higher that you scale. All right, we're out of time, but where can people go to get more information on this and other product, your HCI strategy? If I were them, I'd go to dellemc.com slash HCI. Excellent, Chad. Thanks very much, Stu, appreciate you co-hosting with me and check out the videos on thecube.net. This and other videos will be up there. Thanks for watching everybody. Dave Vellante for Stu Miniman, we'll see you next time.