 Live from Las Vegas, Nevada, it's the Cube, covering EMC World 2015. Brought to you by EMC, Brocade, and VCE. Welcome back, I'm Stu Miniman with Wikibon, joined by my co-host for this segment, Steve Chambers. We'll find all of our research at wikibon.com. Always excited when we have practitioners on the program. Wikibon was founded on allowing IT practitioners to share with their peers. Returning Cube alum, Brian Doherty, Chief Data Warehouse Architector with CMA Consulting. Brian, welcome back. Thank you very much, good to be here. Thank you so much. So we've been digging into the database world. I mean, huge changes going on. Can we remove performance overall from the equation so much? The roles of database administrators versus storage administrators. You've been to EMC World before. What brings you back? Well, what brings me back is all the new exciting announcements that I, especially two that I heard today that are very, very good for us. And that's along the extreme IO enhancements, the 20 terabyte to the 40 terabyte brick and the expansion from the six X brick to the eight X brick. So performance is always critical. What you have to put in to get that performance is really the question. What we're trying to do is to put in less time, but in fact, get greater performance. So they announced it as the beast. The beast. So really the highlight of the keynote this morning, unveiling this, so you've been using extreme IO, gives the update on what's been happening there. Yeah, so extreme IO from the beginning was always a great performer for us. The issue was enough capacity because we support 100 terabyte data warehouses and we have thousands of users and we really needed the capacity also. But extreme IO has gone one step further and taken that 20 terabyte brick to the 40 terabyte brick, which will allow us to get over 200 to 300 terabytes in one extreme IO cluster configuration. Which would be great for us. All right, so I'm wondering if you could go back and talk to us a little bit about what got you into extreme IO, what challenges you were facing before, and really the benefits that you've seen moving to this new generation of architecture. But really one of the first things that drove us to extreme IO was a couple different things. First of all, reducing the storage footprint. We've always needed 10, 16, 20 gigabytes per second of IO throughput for our large Oracle Rack clusters. But we would in the past have a four, six, eight bay storage array to get that. That was a costly storage array. So we wanted something that was a much smaller form factor, still as high performing as we could get on the large storage arrays. And also something we could configure and deploy very, very quickly. So those were the main drivers. Brian, I'm curious, we always often talk what's underplayed in flash is the, it's not just the infrastructure, but the facilities cost that go involved, the space, power, and cooling. Is that something you guys see? Because oftentimes somebody else owns that bill. It's not something that kind of guys to the idea. No, we host our own data centers. So at CMA we also feel that pain. So power, cooling, floor space, are crucial things for us. As we try to scale up into about four petabyte range, those things are very crucial for us. So we're trying to drive more and more performance, larger databases every day, and continue to reduce that footprint. So in terms of performance, we've heard that it changes, it's not just the IOPS, right? It changes the speed of operations as well. We've heard, same as about that. You're a practitioner. Give us some insight into how it changes your operations. So in the past, for the traditional storage array, a lot of our performance architect storage and database would spend literally days or weeks planning a large layout for, let's say, a 50 terabyte, 100 terabyte Oracle Rack warehouse. When Extreme IO came out, now that deployment for us is a matter of five to 10 minutes. So it's literally taken days or weeks out of the equation. And it's also reduced the margin of error. Because traditionally in the past, you have all these different layers that need to be configured, from the RAID type to the device setup, to the front end port setup, to the virtualization setup. And with Extreme IO, all that complexity is abstracted away from you. So it's just much easier today. Because I know EMC, they're often commented as being very good at marketing, aren't they? And I think sometimes something that looks too good to be true, it probably is, and it sounds like, I mean, we saw a demo this morning in the keynote, which was fantastic. They brought an 11-year-old kid on stage and did that, but you're saying, yeah, it's like that in the real world. It really is that simple. It what it does is it takes a lot of the complexity that we had to have separate storage administrators and database administrators, and they were typically separate. And now with Extreme IO and this simplified storage plane, you can really take database administrators or people that are focused on the performance or service level objectives, and they can actually execute the implementation. It's very interesting, and we've heard that there are plugins available. So the Oracle guy from his pane of glass, you can see it. So what does that look like? How does that work? Well, the OEM plugin is out there now. So you can see all of the low-level information on performance. You can see latency on the hours per second, which is important. You can see gigabytes per second that's moving through the Extreme IO array. So you get a lot more, you can see, for example, ASM disk group information. So you can see a lot of performance information through that Oracle window. But surely the DBAs must really miss raising a ticket for the storage guy. They really do. They really do. The DBAs love it. But what about, it's clear that if something's faster for you to do by an exponential amount, you know, it's just so quick to do things. You must be able to do a lot more of them and just change the way you interact with databases and storage. Sure, what you do is you move up the stack. So what we do is we don't spend as, the infrastructure is solid. When you look at an Oracle database performance, the low-level infrastructure has to be there. And then you move up the stack and the database and the physical design and the application. Well, we've taken away a lot of the time and effort at that foundation. If that foundation is not good, you're in trouble. Now with Extreme IO, you know that's going to be good. So we refocus higher up the stack. We can focus on the business. We can focus on the physical design, the application, and get away from the layers that we really shouldn't be focusing on. So Brian, I'm curious. We keep doing a lot of research looking at, you know, even if you spent a little bit more on storage, in the database world, there's huge savings that could mean how to up and down the stack, even including the licensing. Takes a couple of years to sort out some of those things because it's not something that, you know, you turn on demand and everything. Have you done any look-backs on your standpoint as to what the total cost of ownership is after adopting these technologies? We don't have absolute numbers on that total cost of ownership. What we have, though, is we have technology-enabling business practices that we just couldn't do before Extreme IO. So we're running, for example, we have a healthcare cost and quality assurance set of heuristics, a rules engine that we run. And that rules engine scans through billions and billions of rows, and it looks for anomalies, and it looks for healthcare anomalies and cost and quality of care. That would take us days to run before, and a traditional array. Now we can do that in about an hour with Extreme IO. So that's an example where we're into an enabling technology that we couldn't even touch. Well, yeah, I mean, that's transformative to the business, right? That's correct, that's correct. So it allows us to go out and provide kind of a strategic value add to the customers that we service in the healthcare space. Do you see new business opportunities because of this? Do you see yourself doing something different at the path of improving existing stuff, right? When we first bought Extreme IO, we were looking at that really as a point solution for performance problems, specific use cases. Now what we're looking at is putting Extreme IO in as an infrastructure storage layer. So where the use case was very specific to Oracle Rack and Oracle Rack and performance, now we're actually looking at Extreme IO for not only Oracle Rack, but other applications, other databases, other analytics, because some of the enabling technology features here, snapshots, inline-d-duke compression, all of that is really value add to a lot of what we do. So you're looking at becoming the standard. We're looking to build, take it as an extensive of a point solution and build out a flash infrastructure that we can then capitalize on with other applications. So a lot of talk you hear is about matching workloads to infrastructure configurations. Now I know from my past history, I just think, oh my God, that sounds complicated, right? And you can get it wrong. And humans are not very good at estimation of all those good things. Right, so what this does is it gives us more flexibility not to buy a specific technology or a specific storage array and then be boxed in because that storage array isn't flexible enough to scale through a different type of workload. It sounds like a kind of get out of jail free card, just in case someone ever, God forbid you made the wrong decision. That's exactly what, that's a good way to put it. It's really a risk mitigator. You know, when you bring in something that's a small form factor, very good economics, scale out architecture, homogeneous, even performance, great latency, it's really hard to go wrong with that. So it's a great risk mitigator. All right, so Brian, you talked about some of the big announcements that EMC made, but you know, no vendor's perfect. What's on your wish list for the industry in general or maybe EMC product line specifically? What would make your job easier? I think there were two things on our wish list and we got a couple of them today. One of them was the increased density on the brick. So we need petabytes of storage and with the 6X brick and the 20 terabyte, 6 cluster, we were having some capacity issues. But now we got 80, 8 bricks, we've got 40 terabytes. So that's been answered. The second thing was replication. Replication, all of our data centers are hot. So we really didn't have a great replication solution. We had to handle it to the application. But now we're cover point and replication has now come to Extreme IO also. So we got two of the things that were really important to us and I think that's, you know, three years ago when I first started looking at this, we were looking at Flash, we kind of bet the ranch on the technology and the company and they have brought to fruition what we were hoping for three years ago. Do you see yourself using different arrays from different vendors? You must have gone through some comparison and what was your experience of that like? What we saw is a lot of vendors that have very specific Flash products and Flash solutions. What we wanted was a product that really looked just like any other storage array but was really all the performance and benefits of the Flash storage. So for example, we use PowerPath as a multipathing layer for the Oracle Rack cluster. We use that for other arrays and we like it, it works really well for Oracle Rack clusters. We can use PowerPath with Extreme IO. We couldn't use it with other Flash vendors. So there are just things in the Extreme IO package and configuration that were inherited from the legacy of EMC as a company with 30 years of storage industry experience that we could exploit and leverage. As your question about the kind of operation side, there was a day yesterday with EMC talking about DevOps and the impact of different application styles coming into the operations guy and then there's resistance to change and things like that. Did you experience much resistance from any operations or storage guys? Has it changed their job in any good or bad way? Some, but as soon as we got them on the technology and using the technology, it was so easy for them to use. Eventually that went away. Because I assume they've got more things than they can do in a day, right? Yeah, I tell you what it did is it actually allowed them to get much more done in a day. So to a certain extent, it just freed them up to do more work. They was received better by the company internally and they looked good. I mean, Brian, I'm curious. I talked to a service provider who went all Flash and he said, yeah, you know, there's times where I wouldn't necessarily need the performance, but if I can just remove that decision point from what I need to do, boy, it frees me up for a lot. I'm a big believer in we really want that all Flash homogeneous array. I really think that's a great thing. The hybrid arrays are good. They're good for some people, but there's still some decision points that have to go on in terms of what application do I put here? What do I put on Flash? What do I put on non-Flash? And something really nice about having an all Flash array and it's all there. It's simple. Everything's there. It's even performance. It's just really a really nice tool. I mean, clearly you use it for rack. You're running rack-in-it, series-sized databases as well, which I always get nervous with dates, so the fear of losing it, that's just... Early on, there were some issues with the controller. That was like three years ago. We've had it up now running in production, two of them for over a year and we've had no issues. No issues at all. So we've been very happy with the performance, very happy with the reliability, very happy with the simplicity. And the other thing I would say that we really wanted, that I didn't mention before, is we really wanted a very comprehensive command-level interface and that's something that they've delivered in the past couple of releases that really have allowed us to do things, like go in very quickly and cut a snapshot of performance data and then send that into an Excel spreadsheet. Now that sounds like an easy thing, but from a traditional storage array, that's not an easy thing to do. It's a high infrastructure. You have to set up the software, the infrastructure, but with Xtreme I.O., it's all part of the product, the storage management app, and it's very, very easy to get that data. We've also heard that the architecture's different, certainly very close to the array. What about connecting to it? There's lots of different ways to connect to storage. Does it change anything that touches it? Does it change that architecture as well? No, it's really, again, getting back to the point. It's very familiar. It's just like operating where there are other traditional storage arrays. It's fiber channel zoned in, fiber channel zoning, power path there, so it really looks just like our old storage array, but performs a lot better. Great, all right. Well, Brian, we're going to leave it there. Thank you so much for joining. I always love to get your point of view. If you come on another couple of times, we're going to have a whole playlist for the things that we've done. So thanks for that segment. Got lots more wall-to-wall action coming here from EMC World 2015. I'm Stu Miniman with Steve Chambers, and we'll be right back after this quick break.