 Live from the Mendelebe Convention Center in Las Vegas. It's theCUBE, covering VMworld 2016. Brought to you by VMware and its ecosystem sponsors. Now, here's your host, Stu Miniman. Welcome back to theCUBE here at VMworld 2016. Happy to welcome back to the PO program. Christos Caramanalis, who's the fellow in CTO of the VMworld Storage and Availability Business Unit. Thank you for joining us again. Glad to be back. Storage is a big focus here. Big announcements around not only V-SAN, but everything happened in the storage group. Tell us what you've been working on in the last year. Ah yeah, quite a few things. As you know, Virtual Sun has become practically a mainstream product now, especially since we shipped version 6.2 back in March 2016 with a number of new enterprise grade features for space efficiency, new availability features with eraser codes, grade five, six. The product has really taken off. It's taking off, especially in all flask configurations. It's becoming the predominant model that our customers are using. So ultimately, of course, customers buy a new product like V-SAN and hyperconverse product because of the operational efficiencies it brings to their data centers. The way I present this is, you have the operational efficiency of public clouds into your private data center now. But this is, for me, is just the stepping stone for even a longer term, a bolder vision we have around storage and data management. So the last several months now, we have been working on a new range of projects. The main theme there is moving up the stack from storage and the physical infrastructure implications it has to data management and starting with data protection and overall managing the life cycle of your data for protection, for disaster recovery, for archival so that you can have tools to be able to effectively and efficiently discover your data, mine your data, use them by new applications including cloud-native applications. And at the end, even, you know, this may sound a little controversial coming from a VMware CTO, even moving your data to public clouds and allow application mobility freely between private and public clouds. Yeah, it's really interesting. I wonder if you can unpack that a little bit for us. You know, VMware, of course, really dominant in the enterprise data center. We're all trying to understand where VMware fits into the public cloud and how you both support the existing ecosystem and move forward. So, you know, it's an interesting challenge. Yes, of course, there are challenges. There are many open questions. I do not claim that we have the answers to everybody but you see that we put a lot of emphasis on that because it is obvious that the IT world is evolving. Our own customers are gradually, slowly, but certainly they're starting incorporating public clouds into the bigger IT organizations they have. So our goal is to start delivering value to our customers based on clouds, starting with what they have today into the data centers. Let me give you a specific example. In the case of Virtual Sun, we have some really cool tools for managing your infrastructure in a holistic way, compute, networking, and now storage. As part of that, we have solutions and tools that allow the customer to monitor constantly their hardware infrastructure, the configuration of that, the clusters, the network, the servers, controllers, down to individual devices. And we provide a lot of data to the customers, not only for the health, but also for the performance of the infrastructure. Data that the customer can today use to perform root cause analysis of potential issues to decide how to optimize their infrastructure in the workloads. But that is actually a pretty sophisticated task. We cannot expect all the 500,000 customers of VMware to be able to do this kind of sophisticated analysis. So what we're working on right now is a set of analytics tools that do all this data crunching and analysis, a root cause analysis and evaluation of the infrastructure on behalf of the customer. And instead of providing data, now we're providing answers and suggestions. Now we want to be able to deliver those analytics in a very rapid cadence. So what we do is we develop all those things in VMware's cloud. We collect data from the customer side through telemetry, VMware's phone home product. And we get all the data up in our cloud. We crunch the data on behalf of the customer and we use really sophisticated methods that will be evolving over time and eventually we'll be delivering feedback and suggestions at a high level to the customer that can be actionable. For example, we can point out that certain firmware, firmware of certain controllers in the infrastructure is falling behind and may have problems or point out to certain SSDs getting close to the end of life or more sophisticated things such as reconfigure your application with a different policy for data distribution to achieve better performance. The interesting thing is that we're going to be combining data from multiple sites and multiple customers to be able to do this holistic analytics and say, you know what, based on trends I see on other customers, I suggest you also do that. Now the really cool thing out of this is that the customer does not have to go and use yet another portal on a public cloud to take advantage of that. But in fact, we send all that feedback through the vSphere UI on-premise to the customer. So really cool. You have the best of both worlds. The rapid development of analytics using actually, I'm behind the scenes, is a really complex cloud native application with the existing tools that the customers are used to on-premise. So this is just one example. Are you set? Christos, could you give us a little bit of insight as to the guiding light for your development process? Is it kind of core customers that you're pulling in and working in? Is it a mandate from above that says, you know, hey, we need to build a more robust and move up the stack? You know, what are some of the pieces that lead to the development that you're talking about? This is a very interesting point. I must start by stating that VMware has always been an engineer-driven company. A lot of our products were ideas that were, you know, imagined by engineers while others thought that was not even feasible. And of course virtualization in its early stages, but features like VMotion or storage VMotion or, you know, even, you know, ideas behind the virtual sun, right? Claiming that I could do very effectively RAID 6 in software was something that was not really, you know, appreciated in the industry at earlier stages. So a lot of the innovation is grassroots innovation. We have our engineers exposed directly to customers, to customer problems. Of course, they also understand what is happening in the industry, the trends, whether that is in hardware, as it's the case these days with a new generation of storage hardware that is emerging, or whether that is a trend among customers, for example, using public clouds in certain ways, whether that is for doing test and dev or archiving their data. We observe those things. And then through a grassroots effort, all this get amalgamated into some concrete ideas. I'm not saying that all those ideas result into products, but we definitely have a very open mind in letting engineers experiment and prove sometimes a common sense to be wrong. So this is the process. This is how virtual sun started. Where a handful of us went to our CEO back then for marriage and suggested we do this drastic thing that is called, you know, softer storage and that you can run the storage stack in software on the same servers that we virtualize and we run the VM. So this is really how the process has always been working and this is still the case. And we're very proud of this culture. This is one way we're actually attracting top engineering talent in the company. Yeah, I always loved digging into some of the innovation processes. Had a good chat with Steve Herrod, former CTO of VMware. If I remember right, one of the thing processes you use are called flings, whereas you can actually get visibility from the outside into some of those kind of trials and things that are going on that aren't yet fully supported yet. Yeah, absolutely. And that is still the case. Probably the best known fling these days is the HTML5 based UI for ESX, which is used extensively. Both internally in the VMware where we actually started as a tool for that purpose but now widely by the community. And that fling gave us a lot of insights in how to evolve our mainstream user interface for vSphere proper, not just for ESX. So this is exactly this iterative process that leads us to test the water and feel much more confident when we make bigger investments in R&D. All right. Architecturally, VMware's been around for quite a while now. I had a good talk with Sachin Vagani, who created VMFS earlier today. And we were talking about new applications and new architectures. When VMFS was built, nobody's thinking about containers. They weren't thinking about applications like Hadoop or some of these more cloud native applications. How do you take into consideration where things are going? How do these fit into kind of traditional VMware vSphere or what things need to change? How do you look at kind of the code base itself? Right, right. So first of all, VMFS, I must say, it's probably the most mature and most widely adopted cluster file system in the industry. For over 10 years now, it's been used to virtualize enterprise-grade storage. Storage-area networks. And it was going to have a role for many years to come. But on the other hand, we already are, you know, technologists and we understand that the product is designed with certain assumptions and constraints. And VMFS was designed back in the mid 2000s to address the requirements for ritualizing lands and, you know, the traditional volumes that you'd be consuming from a disk arrays. Now the world is changing, right? We have a whole new generation of solid state devices for storage. Servers and software on commodity servers with commodity storage devices is becoming, as your own reports have been indicating, the predominant model of delivering storage in the enterprise data center. And of course, in even public clouds with hyperscale storage. So what the requirements, the assumptions are changing. You need the storage platform that can really take advantage of the very low latencies of those devices. I was at Intel developer forum a couple of weeks ago and their Intel announced for first time performance numbers for the new generation of NVMe devices obtained that include the 3D cross-point technology under the covers. Latencies at around 10 microseconds, right? And Ios per second, you know, throughputs that are in the several hundreds of thousands if not millions. So completely, you know, game changer. And Intel is not the only company that is coming up with this technology. So you need to invest now in new technologies that can take, that can harness the capabilities of these new devices. Lightweight protocols, like NVMe. In fact, I see NVMe as a protocol is not just a protocol to access the device but they can see a future for that of replacing SCSI into the software stack. Soon, and this is, you know, commit to any specific dates but soon we'll be shipping a version of vSphere that comes with virtual NVMe in the guest, virtualization of NVMe. So you can see here where we're heading and NVMe becoming a predominant protocol for the transport and for virtualizing storage. Yeah, interesting. And we've got a long history of things that start out in the guest. It usually then takes a lot of engineering work to get them down to the hypervisor themselves. So, you know, without having to give away too much is that we see that kind of progression sometime in the future for some of these new memory architectures? Oh, yeah, certainly, certainly. The ESX storage stack. And this is the stack that is used by VMFS, by Virtual Sun. It has been designed again for another era of storage. Now we are regarding a lot of these things there and I cannot disclose too much detail obviously but I can tell that it's going to be a very different software stack, much leaner, much more optimized for local, very fast devices. And ultimately, NVMe is going to be a key technology in this new storage stack. All right, so just last follow up on that topic. I think about kind of the new memory architectures, what's going on. As of September 7th, Dell will have acquired EMC. There's the relationship between EMC and VMware. Could we expect some of these new memory technologies and packaging things to be something that you'll work even closer with Dell EMC? That is definitely the case, irrespective of the deal between EMC and Dell, which as you said is going to be closing, it seems pretty soon from what I read on the newspapers. Michael confirmed it, so it will be official. We can move beyond some of the hypotheticals. Yes, we're moving ahead with these new technologies and we're working closely with all the partners, Micron, Intel, and many of the other hardware vendors that are introducing such technologies to incorporate them into our systems, into our software. For example, I see great opportunities for this very fast, high endurance, but still quite expensive technologies to be used, for example, to store metadata. Things like the duplication host tables, those kind of metadata that have an impact because of IO amplification, to the performance that is perceived by the application. By moving metadata like that into those tiers, I'm going to make a great difference in terms of performance, consistent latency, and predictability at the end of the day for the application. Now, thanks to the relations with Dell and EMC, I can hope that some of these technologies will find their way into server platforms sooner than later, so all of us and our customers will benefit from that. All right, well, Christos, really appreciate getting the update from you, lots happening on the storage world. We're kind of talking about one of my things coming into this week was, if we can really simplify storage, we might actually have a storageless world. Doesn't mean it reduces the value of storage or the importance of it, but going to help the users to be able to move beyond that. We'll be back with lots more coverage here from VMworld 2016. You're watching theCUBE. 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. Now, here's your host, Stu Miniman. Welcome back to theCUBE here at VMworld 2016. Happy to welcome back to the program, Christos Caramonalis, who's the fellow in CTO of the VMware Storage and Availability Business Unit. Thank you for joining us again. Glad to be back.