 And from Las Vegas, it's theCUBE. Covering EMC World 2016. Brought to you by EMC. Now, here are your hosts, John Furrier and Dave Vellante. Okay, welcome back everyone. We are live in Las Vegas for EMC World 2016. This is SiliconANGLE Media's theCUBE, our flagship program. We go out to the events and extract the signal from the noise. I'm John Furrier, my co-host Dave Vellante. Our next guest is John McCool. SVP and General Manager of DSSD at EMC. Welcome to theCUBE. Good to see you. Thanks, John. Obviously DSSD is the darling in flash. So how's it doing? How's it going? I asked this to CJ. How's it going? It's great. You know, if you remember this time last year, we were introducing the technology. We talked about face melding performance. And since we released in March, we've been finding out what face melding means to customers. So data intensive applications, both existing and new are developing fast. David Floyer from Wikibon was earlier on our kickoff segment, really talking about the need to reduce the latency and make things faster on both sides of traditional and kind of next generation apps. Whether it's moving compute to the data, advice for survival. A lot of different approaches, depending upon the architecture. What are you guys doing there? And what are some of the trends in the industry is, is that recognized that that's a need? And I know DSSD has some advantages there. What is that dynamic for that, that data intensive environment? Yeah, I mean, it's really around analytics. So people trying to do fraud detection. And if you think about a bank, you put your credit card in, the time to detect security and fraud is in the duration of that transaction. So how do you minimize the time to detect a threat and react to it? So it takes a different manifestation in every vertical, but that's what it's all about. And what are the next generation apps? What did they look like? Yeah, so I mean, we've seen some interesting trends. There's folks that have their traditional databases and they're trying to extend them to do things around analytics and extract the information they already have. And they're doing some very contrived things with multiple views and indexes and their environment in the application sense is getting pretty darn complicated. The same time, they're trying to move over to things like Postgres and HDFS to kind of minimize the pressure on those traditional applications. So we see rack scale as a way to bridge from an infrastructure perspective, both of those environments. So I wonder if we could do a little DSSD 101 for people who aren't as familiar. First of all, what does DSSD stand for? That is a mystery that even I don't know. SSD, obviously, the D is the mystery one, okay? So thinking about the way in which applications are developed, there's always the assumption that when I do an IO, I'm going to go through this horrible storage stack and I'm going to go to some spinning disk and I'm going to wait and I'm going to do other stuff while I'm waiting for them. And you go out over a network of fiber channels. It's the only mechanical piece left and of course with Flash that sort of changes. But that's how applications have been written for decades, half a century. Okay, so what problem are you solving with DSSD? Well, there was an alternate. So you could build out a Flash array and put it on a fiber channel network and when you want it more throughput, you would scale that fabric out, which is great and it works to some extent. But at some point, you're over provisioning. In that work, you're over provisioning the storage just to get that throughput. And you're still going through a chatty, scuzzy protocol or whatever protocol you're using. The alternate approach was to directly touch that drive into the compute itself and you see people building foresocket machines and lots of memory and beefing that up to try to run their single application. The data is then captured and can't be shared. So we're bridging PCIe directly connected. So we're bridging the concept of the old Flash array, but still directly connecting it through PCIe into the server. So we become a piece of that compute infrastructure almost more so than the storage element. And you're cutting the line on the protocol, right? The storage protocol, right? The NVMe protocol goes directly to a drive today inside compute. We're just extending that outside using the modern protocols that all compute talks today. So it's near memory speeds, memory extension that's persistent, right? So we're talking about multiple orders of magnitude of performance increase, right? Yeah, you're taking milliseconds to 100 microseconds in terms of all the insane. Now my understanding is that in the early days of FusionIO, we loved the vision. There were some challenges, right? You had to rewrite the applications really to take advantage of it fully and you mentioned the sharing challenge, but the vision was laid out. You guys are actually delivering on that vision and beyond, right? We're doing a couple things. I mean, we have a traditional block driver that we've extended for performance so you can run existing applications on a file system directly on the product. We've taken our own API and we've built a plug-in into HDFS in conjunction with cloud era. So if you have an HDFS infrastructure, you can directly port that onto a D5. And then we have a lot of customers specifically in the financial services area, DOD, Defense, looking at writing directly to the API and custom applications that they have today. Okay, and so the business impact is massive game-changing productivity, doing things that you couldn't do before. Maybe you could give us some examples or describe some of the things that you're accustomed to. Sure, I talked about fraud detection at Genomics. We have companies looking at taking the time and I think we talked about that where every child will have a profile and their DNA in the future. I mean, we're taking the time down for DNA genomic sequencing down considerably. So these kind of things. So take fraud detection, for example. I mean, one of the challenges in fraud detection is there's still too many false positives, right? Are you guys attacking that problem with your customers? Yeah, so you can take more data and to get better results as a result. And also take the time. We've talked to a lot of customers who are doing this in batch mode. So by the time a detection of a fraud occurs, you've left or somebody has gone. You can't capture the actual event and correlate it in time. So we can make that real-time information that's available at the point of sale or the time something is processed. John, I want to just get the announcements out of the way because I know you guys had some slew of announcements. Just share with the folks. Take a minute to share what you guys have announced here that the show this year. Yeah, so we released the product. We announced it publicly on the 29th of February. We started working with customers specifically in the area of high-performance databases. And they noted, you know, with a drive that's inside a compute farm, I can actually stripe across that. I should be able to do that with D5. Can I do that? And it actually works. And it was just built into the product. And we productized that and announced a single client can connect to two D5s so we can double the performance on an application like Oracle. We had benchmarks. We announced a 5.3 million 8K IOPS at the launch. We now can do 10.6 with the same latency. So really breakthrough performance. We also announced a CI configuration with VCE in a VX rack. So this will become preconfigured with a D5 so people now can consume at the rack level. And we think that's going to be very significant. People who built by the compute infrastructure connected in with the D5 directly. What's the customer reactions? But I know you guys had been doing a lot of on the beta side, pre-GA and some of the announcements. Can you share some color around like, are people falling out of their chair? Or are they like, oh my God, a huge increase in preventing X or generating more revenue? Can you share some color around some of the anecdotal feedback? It's really interesting. So there's a whole technology appeal to this and in some ways kind of an obvious aspect to the technologist of yeah, it makes sense. You connect the compute directly to storage. 10 million IOPS, that's unbelievable. Like what do I do with it? And my traditional infrastructure has been designed in a certain way. And then sometimes we'll see an aha moment of something they regret it not doing in the past. Don't you remember, we had one CTO say don't you remember we had this application and then we over provisioned and built this network and had all this storage. And we could have done that with this today. So it's not necessarily things in the future but regrets in the past that- So they're seeing instant low hanging fruit that they can go after. And that goes right into the simple business case. So help us understand, we talked about the insufficiency of traditional storage. What's the price premium? If I had a hundred dollars, if it costs a hundred dollars for traditional storage, what's the price premium on percentage terms that I would have to pay roughly for a decent thing? Yeah, so on a raw gigabyte basis, we're compatible with today's all flash arrays. So no issue in terms of comparison on a raw gigabyte per second basis. So really no price premium. Exactly. You know, obviously we have a dense box and we can do a lot of compute around it but on a capacity basis, we're very compatible. So let me make sure I understand it. You're adding additional value besides obviously just the capacity that I would pay for as a customer. It's different value. I mean, when you have an all flash array, you can connect into a legacy network. You have iSCSI, you have multiple protocols, you work with VMware. We're really based on transaction applications. I mean applications around analytics and high speed, ones where the data tends to be refreshed very often and a lot of processing behind it. Right, right, okay. So when you talk about the customer saying, hey, I got all these, I asked what I do with it. Obviously there's probably some instant areas but then they get the aha moment of, they start dreaming about things they couldn't do before, as you mentioned. As a product team, how do you guys look at the roadmap? And what do you say to the customers when they say, hey, I love it, I'll start working on this immediately. Thank you very much for the product. But what's next? Where's the headroom? What are you guys, how do you talk to the customer and give them that confidence going forward? What's the vision, what's the guiding principles? I mean, you don't have to give specifics, but I mean. Yeah, I mean, there's sort of two dimensions to this. There's a horizontal dimension that's typical in all things, bigger, better, faster. And obviously we pursue that. But more and more solutions targeting and application specific needs and understanding those applications and workloads and figuring out how we can migrate them in the future. How can we help you with your database problem today but move you to these modern technologies like Hadoop? And how do you do that migration with a single piece of infrastructure? So a lot of new capabilities basically are evolving for them. Correct, correct. All right, John, we appreciate you taking the time here. I'll give you the final word on the segment. For the folks watching who couldn't make it here, what are the things that should jump out the most that they should pay attention to with your announcements and your demos and your presentations this week? Sure, I guess I talked to the CTO audience and some of these enterprises and remember those things you couldn't do before and dust them off and come talk to EMC about what you might be able to do in the future. All right, John McClure. SVP and General Manager of DSSD, the hot flash darling at EMC and doing great. Thanks for sharing your insights here in theCUBE. We'll be right back with more blazing performance from theCUBE after this short break. I'm John Furrier with Dave Vellante. You're watching theCUBE. Looking back at the history of Dell, personal computers.