 Live from San Francisco, celebrating 10 years of high-tech coverage, it's theCUBE. Covering VMworld 2019. Brought to you by VMware and its ecosystem partners. Welcome back, we're here at Moscone North for VMworld 2019 10th year of theCUBE, covering VMworld. I'm Stu Miniman, my co-host is John Troyer and welcome to the program. Two guests from Dell Technologies, sitting to my right is Chinumai Mondal, who's the director of storage solutions and sitting to his right is David Wen, who's senior director of server product planning and management, also with Dell, gentlemen. Thanks so much for joining us. Thanks for having us. All right, so we've got server and storage. I'm going to talk about something that we've been talking about for a while on the server side, it's been delivered for a bit and on the storage side is now rolling out. So, everybody's favorite topic, non-violatal memory express or NVMe as it rolls off the tongue, storage class memory or SCM and lots of other things down that are really helping, a big transformational wave that really changes how our applications interact with the infrastructure. Chinumai, bring us up to date on the latest. Sure, and let's start where you ended. We are seeing explosion of applications, right? And in fact, in Morning's keynote, Pat Gelsinger had a statistics, there are 352 million enterprise applications today and it will be 792 million in three years. Now, as the applications are growing exponentially, we cannot keep growing the infrastructure at that rate. So, NVMe is the way we can consolidate a lot of the infrastructure. If we can think about end-to-end NVMe starting from the server NVMe over fabric to the storage array down to the back end with NVMe SSDs, this actually can put together a great platform where you can consolidate a lot of the applications and delivering the high performance low latency that we'll need while meeting various surface level objectives. So, we can go over a little bit of the details, but I think it all starts from NVMe over fabric coming from the server to the storage array. So, probably that's the first step we need to consider. David, I love this discussion when we get to talk at the application later because Flash changed the market a lot. It's like much better energy and it's much faster anything, but this inflection point that we're talking about for application modernization, NVMe is one of those enablers there and something I know your team's been working on for a while. Yeah, actually on the PowerEdge side, we've been embracing the benefits of NVMe for quite so many years now, right? We started out by introducing NVMe in our 12th generation servers, front-loaded hot serviceable drives, and then of course we branch out from there and then today a lot of the servers from the PowerEdge family all support NVMe devices. So, the benefit there is really given customer choices in terms of what kind of storage, what kind of tiering they want for the application's needs. Now, one of the things that's great about NVMe over fabric is it's more than just the Flash storage itself. It's about enabling the standards across the data fabric down to the storage really to deliver on the overall performance that the application needs and by improving IOPS and lower latency overall from a server perspective, this just means that we're releasing more CPU cycles back into the application so that it can run different types of workloads. And for us, this is a great story from PowerEdge as well as from PowerMax coming together to enable this NVMe over fabric. You know, I'm kind of slow about some of these things but if you kind of squint at the history. And we went from the PC revolution and then we had sands and arrays, right? And then we had centralized storage, shared storage. Last couple of years, a lot of interest and still, right, hyper-converged and you had a lot of pizza boxes with the storage right there. I mean, I now think, right? And I'm following the thread, I think, which is now that we now can have a rack with again a fabric and again now we can focus on our NVMe over fabric driven solid-state storage somewhere below my servers that are doing handling compute somewhere else. Is that the future we're headed towards now? Yes, I mean, everything has its place but to give you the perspective, right? It's not just, I mean, coming down to the storage array but how this is enabling the future storage as well and the storage class memory is the perfect example. And as David said, let's take PowerMax as an example, right? So in PowerMax, it is like end-to-end NVMe ready like you get NVMe over fabric at the front end but then we have NVMe SSDs in the back end. The thing is now it is also the NVMe is enabling technologies like storage class memory which is bringing in like very high performance, very less latencies, latencies going down in the order of like tens of microseconds. Now, this is as close as you can get to the like DRAM with persistent storage. However, you need a balance. This is like order of magnitude costlier. Now look at PowerMax, what we are doing. In terms of first, it's NVMe done right. What do I mean by that? You have like multi-controller architecture that can actually do this level of parallel processing and concurrency and then we have both like SCM or storage class memory and NVMe SSDs and we are doing intelligent tiering based on the built-in machine learning engine that we have and it is looking at 40 million data sets real-time to decide like which set of workloads should go on this SCM drives which should go on NVMe SSDs and on top of it, you add quality of service. So this platform gives you service level objectives. You can choose from diamond, platinum, gold, silver, bronze and you can consolidate a lot of those 352 million different types of applications on this array guaranteeing you are going to meet all of your SLS no matter what type of applications you are consolidating to. Okay, I'm wondering if you could both, you know, bring us into what this means for VMware customers and break it into two pieces. One is kind of a traditional virtualized shop and secondly, you know, spent a lot of time in the keynote this morning talking about the cloud native containerized, you know, type of environments. Will there be any difference from both of your worlds? Yeah, absolutely. I'm glad you brought that up because, you know, from our perspective, right? What we've seen with the enablement of NVMe in our platforms, you know, John, you brought up a very interesting point, right? It seems like, you know, it passed a couple of years we went from moving storage onto the host and now with NVMe over fabric, we're actually taking the storage away from the host again, right? And that's exactly true because, you know, the first statement you brought up still is about how Flash enabled different applications to run better on the host. Well, we see that still, right? And so with NVMe, you know, we see the lower response time enabling our customers to run more jobs or more VMs per server. That's one aspect of it. You know, we've seen this benefit. A lot of our platforms today are using various different applications and solutions. You know, you talk about VxRail. That's a VSAN story for Dell. You talk about VSAN ready notes for customers who want to build it themselves, right? Our platforms enable with NVMe backplane, NVMe storage, allows them to use NVMe or Sassassada, whatever they want. But the point is here is that when they're using NVMe Flash, for instance, and I'll talk a little bit about the PowerHMX with this all Flash NVMe backplane. In a case, in a study that we did with VSAN application running an OLTP type of workload, we saw the response time with NVMe over traditional SaaS from our competitor improved by 56%, right? Which means that from that same particular solution build out, we were able to add 44% more VMs on the platform. Now, at the same time, we increased the overall orders per minute by roughly over 600,000 OPMs for that type of benchmark over our nearest competitor. So that right there is the benefit that we see from a virtualized, from a VMware perspective. And I'll add from the storage perspective in two ways. In fact, in last VM world, in EMEA, we demonstrated end-to-end NVMe over fabric with special build of this sphere supporting NVMe over fabric and storage class memory with NVMe drives. What it gives a regular like the sphere-based environment is that you have the ability to move your VMs around like the applications where the highest performance and latency is critical. It will be on those special service levels and special like data stores. In fact, the demonstration was like SCM data store and NVMe SSD data stores. So in the same fabric with, in PowerMax, you can move things around whether it's like regular fiber channel or SCM. And then the other part I want to add in the morning, like we saw the announcement that now Kubernetes is built in or will be built in with the ESXi platform, right? And ESXi is bread and butter of all the storage customers that we have. Now, when you consider those things built in on the vSphere platform, think about like how many applications, how many virtualized workloads you can run, whether it's on-premise or VMR cloud on AWS. All of those consolidation as well as like the performance needs while reducing your footprint, that's the benefit the VMR shops or the VMR admins are going to see from the storage side. And again, I'm not following the parts, but what kind of, I mean we're not talking about a couple of megabytes here anymore, right? What size of parts are shipping these days? How much? So from our perspective, up to seven gigabyte, actually I'm sorry, seven terabyte drives are available in the market today for NVMe. Now, whether a customer buy those drives, you know, it depends on the economic factor. But yeah, it's something that's available from Dell today. So, and I'll add to what David said. So for SCM drives, 750 gig to 1.5 terabyte SCM drives, dual-ported opt-in drives that will be available in the PowerMax as well as 15 terabyte NVMe SSDs. So this is the capacity we are talking about. And again, the latency is at the application level, like from the storage, like you are going to see like less than 300 microseconds. That's the power we are bringing in with this technology to the market. Great. Give us a little look forward. We talked about, you know, NVMe's been shipping for a bit on the servers, now it's really rolling out on the storage side. Saw there's a lot of startups in the space, you know, when recent acquisition got some people talking, what should we be looking for from both of you over kind of the next six to 12 months? So over next six to 12 months, you will see a lot of innovation in this space from the storage side where the order of magnitude. The one single array, I mean today it supports say like 10 million IOPS less than 500 microsecond latency. I cannot give you the exact details, but within like a short time, these numbers are going to go up by like more than like 50% like latency is going to get reduced. The throughput will be driving will actually like more than double. So you'll see like a lot of these innovations in kind of like evolution in terms of the drive capacities both from the SCM drives perspective NVMe SSDs, those will continue to expand leading to faster performance, better consolidation for all the workloads. Yeah, from our perspective, I mean, you know, the data growth is going to continue. We all know that. And for us, it's like designing systems based on what the customers need, what the applications needs, right? And hence that's why we have different types of storage that's available today. So for us, you know, while we're doing a lot of things from a direct attached storage perspective, customers continue to have a need for shared storage. NVMe over fabric just provides a better, you know, end-to-end story for us really from a PowerEdge and PowerMax perspective. But in the future, you ask what we're going to do. Well, we see the need to probably decouple storage class memory from the host again. And really what's preventing us from doing it today is really having the right fabric in place to be able to deliver to that performance level that applications needs. NVMe over fabrics, fiber channel, Ethernet, iSCSI, or I'm sorry, Infiniband, whatever. These are some of the things that, you know, we're looking forward to in the future to make that leap. All right, well, it's really been great to see to the technology that I know the people that build your products have been excited about for many years, but rolling out into, you know, real world deployments for customers that will transform what they're doing. So for John Troyer, I'm Stu Miniman, back with lots more coverage here from VMworld 2019. Thanks for watching theCUBE.