 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. Hey, welcome back everyone. We are live here in Las Vegas, this is Silicon Angles, theCUBE, this is 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 Octavian Rotar, who's the lead technical architect at Amdox. Welcome to theCUBE. Thank you. So, you're lead architect, so you're under a lot of pressure. So what's the bottom line? What is this cloud about? What are people doing? What's working? What is the things that's working the best? Well, depends on your requirement, depends on your application, your application makes. I think we don't have a silver bullet today. And the silver bullet, it's not cloud, it's not private cloud or hybrid cloud or owning your own infrastructure. You have to adjust what you do and your implementation to the specific requirements of every customer. So let's step back, this is kind of a general question, but this is a conversation that many people who are in a traditional environment or are in cloud native mode, which is DevOps, are looking at architectural stuff and making some new decisions or at least thinking about what to think about and how to architect. So a lot of people are having these conversations because beauty's in the eye of the beholder, meaning your own workflow or apps have to drive a lot of the input into the engineering. So take us through the process of that. What are you guys doing? How are you guys rolling out your tech? And how do you think about your architecture? Okay, so what myself and my team are doing, we are, usually we are receiving a set of requirements that we have to meet. And what we do it is to translate that application mix, that set of requirements for the customer into an architecture, into a technical architecture that starts from computing power. Servers goes down all the way to network, storage, sand infrastructure, and everything else related. So in order to do that, what we have to keep in mind that you know, every time you have to write size, because if I'm going too low and I'm undersizing my system and I don't have what it needs, I'm not going to meet my SLAs. My customers will not be happy. If I'm going too high, then I'm going in the other side of the story. First of all, if my customer can say, why did I buy this? Why do I have more than what I need? So I always have to be on the, try to find the sweet spot between safety for your SLAs and meeting the requirements. And in the same time, what is finding the right spot for your hardware. And you got to deliver the value because you always got to be on the edge of getting better performance. So, you know, all flash has been a really big part of the strategy EMC. How are you guys using flash right now? Because that's an instant game changer in terms of performance at least. How do you bring that in? And what workloads do you decide? Or all workloads, what's your thoughts on all flash? When I'm looking at all flash, I'm seeing it as a technology enabler for our application workload. Because we are working in the telecom market and providing services and software for telecommunication providers mainly. And the type of workload that we see is usually high density workload, requires a lot of IO. And we are looking at all flashes of technology enabler to meet our SLAs, to meet the requirements from our customers. And in the same time, achieve efficiency. And I'm not talking here only about price, also data efficiency. Because we have a lot of, in most of the environments, we have a lot of copies of the data. You have the primary copy of data and then you take copies of it for development, for performance testing, for backup. And on all flash, you can reduce the footprint of data because you are going to run mainly snapshots. You get the right level of performance, even with so many copies. So you're a long-term symmetric customer, right? Yes, we are a long-term symmetric customer. We went a long way, starting with all the old arrays, traditional arrays with mechanical spindles. Then went to hybrid storage arrays, tearing fast. And today we are reaching a point in which all flash makes more sense. So as a VMAX now, symmetric customer, how do you look at the platform? Does the all flash, the announcement of all flash, is that a game changer for you? Does that make it more strategic? Is it more of a tactical solution? What do you say to that? In our case, I don't think that this is a game changer today. Why? We started to use VMAX arrays in all flash configuration before this was productized by EMC. We have right now VMAX arrays in all flash. I don't think that we are the one to show them the way, but I'm sure that other customers requested that as well. So we already have in production for over one year the VMAX arrays that are running 100% flash. VMAX 400K in that case. And we did that because we have certain workloads that require very, very low latency and predictability. I need my response time to be always the same. I have to stay in a very low range. The moment I'm tearing such an application, such a database, then my response time on average will be okay. But some parts of the data will have higher response time, usually the ones that will be tear down. And this can create issues downstream. And at the end of the day, what we are talking about is user experience. And usually when you use your phone, or when you use an app, when you try to make a call to text somebody, you don't have the patience to wait three seconds or whatever it takes for an application to answer down the road. You want your response immediately. Most of the times, if you are opening a webpage today or an application, if it takes more than couple of seconds to respond, I think that most of the people will close it. So okay, so it's not a game changer because you guys had experience beforehand. But if you look at, I can't remember when it was now, 2009 maybe, when EMC first put flash drives inside of Sybetrix, I believe at the time. Yes, inside of Vmox. What did you know of Vmox? Yeah, it was Vmox One. Right, and so that definitely elongated the viability and the value contribution of the platform, no question. EMC was first there. And now, fast forward to 2016, you have this all flash configuration. How important is it for M-Docs, generally, and you specifically as the lead architect, that you have that stack in place and you're able to leverage that? Because, you know, a lot of people say, okay, well, Vmox, or do I go with ExtremeIO? But the stacks are different, right? The functionality is different. Why Vmox for you? Okay, if I'm looking at Vmox and ExtremeIO, they cover different requirements for a couple of reasons. If you are looking for remote replication of data, storage-based replication, then the best option would be to stay on Symetrix. Also, an additional reason for us, it is that internally we have a lot of tools that are developed and they are based on SimCLI, on Symetrix CLI. The moment I have to implement ExtremeIO, I do not have the same tools available today. One year from now, maybe I will have them, it will be different. So, the moment I need tight integration into my environment and I need to use existing tools that are based on SimCLI, if I need storage-based replication to a remote site, all these things are better addressed by Vmox. Also, if I'm looking at the type of workload, if you have a big oracle database, a massive instance, then I would probably look at it more on to put it on Vmox, not on ExtremeIO. In the same time, if I'm looking at ExtremeIO as a use case for me, the best that I can do is virtualized environments, lot of VMs, repetitive data and so on. The moment you run an oracle, a big database, a big oracle database on ExtremeIO and not only ExtremeIO. In any old flash array that is doing the online compression and data duplication, yes, you will benefit from compression. You will also benefit from the duplication, but very, very little. The ratio of the duplication on oracle databases tends to be very small. You start to improve your data efficiency only in the moment when you have additional copies of the data, but until that point, your data efficiency is quite low and I prefer to run such a workload, especially when the IO size tends to be bigger than normal on a Vmox, on a Symetrix array. And that's a cost factor. You can't exploit that data reduction, you're saying. Okay, and so my understanding is that you brought in the Vmox All Flash for internal purposes, internal apps, and then after you got some experience there, you started to support your external customer-facing apps. Is that correct? Did I get that right? What we are doing, so first of all, my team works for different projects. Some of them are services, managed services projects in which Amdox is owning the hardware. We run the business and offer the services to the customer. But in the same time, we also provide architecture for external customers, in which case they will have their own hardware solution. We may implement it or they will implement it or somebody, a system integrator, may do it. So initially, when we started with All Flash, we started to run our application in-house internally. This is part of the managed services business that we are doing for certain telecommunication providers. And then we started to roll out and to propose all Flash architectures, of course, to customers that own their own infrastructure. What did you learn from that transition from internal only to customer-facing? Were there learnings that you could apply? First of all, what I noticed in the past one year or so that we are being asked more and more by customers to provide a solution based on All Flash. So it's not only coming from our side, but from our side. But the customers are also embracing it. We have many customers that they are asking for it. They want to go forward. They want to innovate in technology, to have data efficiency, to have predictable response time and the solution that they know they can grow into for a long time. Octavian, thanks so much for joining us on theCUBE and sharing your insight today. If I could give you the final word. Are you excited about Flash? Yes, I am. Everybody is? Yes. Is spinning disk going to go the way of tape? I think it will. It will take a while, but it will. Spitting rust, as Dave would say. Well, we love the Flash, seeing the performance. Thanks for sharing the insights here on theCUBE. I'm John Furrier, Dave Vellante. You're watching theCUBE at EMC World 2016. Looking back at the history of Dell,