 Live from the Moscone Convention Center in San Francisco, California, it's the cue at Oracle Open World 2014 brought to you by headline sponsor Cisco Systems with support from NetApp. And now here are your hosts, John Furrier and Jeff Frick. Okay, welcome back everyone. We are live in San Francisco, California, the Moscone Center for Oracle Open World 2014. This is theCUBE, our flagship program. We go out for the events and expect to see the noise. I'm John Furrier, the co-founder of SiliconANGLE Media and I'm here with two distinguished guests from Cisco Systems in the IT space. We have Ravi Amamatu, who's an IT manager and we have Hari Shankar, technical staff, global infrastructure services at Cisco. Welcome to theCUBE. Thank you. We appreciate you guys coming on because we talked to all the Cisco executives. They give us a lot of Cisco this, Cisco that, Cisco's great, value proposition. We love to talk to people in the trenches. Obviously we just had JP on earlier talking about Oracle and Cisco. So you guys have been joint customers for years and Cisco has founded Silicon Valley Connection. But I want to ask you guys about your IT experience. You can share the folks out here watching. What were the lessons you learned in IT? I mean, Cisco is very progressive and very innovative and they eat their own dog food or whatever analogy you want to use. They're on the bleeding edge but they're also very high quality, secure. I mean, I tried to get a white paper. I couldn't get in with firewall. I mean, it's network gear. So you guys have one of the most established IT practices. So what's the experience working there? What's it like working at Cisco? Yeah, so, you know, if you look at the industry, right, you know, it's very challenging as well as the, you know, industry is moving very fast and it was very difficult to cope with the change that's happening. And the amount of change happening with the amount of data, right, any IT people need to handle today is phenomenal, right? And when the change in the data is happening, probably one of the challenges basically, how do we fast provision the servers? How do we, you know, provision the servers fast? Plus, how do you, you know, both scale vertically and horizontally to meet the business needs? At the same time, one of the biggest things that we need to take into account is the performance of the applications itself. You know, while we are taking care of all these things to account, how do you reduce your TCO, basically the total cost of, you know, operations to account, right? I think the IT budget, as we all know, it's reducing year by year. And one of the challenging, you know, factor as an IT manager for me is that how do you scale your environment better with the reduced cost year over year? Yeah, they want you to do more, though. Hold on, there's a trick. No more budget, lower budget, more functionality. And how do you do it? And that's exactly the UCS comes into play, right? We are running, actually, Oracle databases on UCS. And when we are looking at... Hold on, hold on, before you go there, I want to just stop for a second. So let's just pause right there. So do more with less. That's a phrase we've all heard before. So let's talk about the Oracle situation. So you have Oracle, huge shop, obviously the relationship with Oracle. But now, you migrate Oracle to UCS. Take me through play by play. Why you did that? And what were some of the things you did? What did you learn? And what were the results? Yeah, absolutely right, John. So before I go there, right, maybe I will give you a bigger picture of how many environments we run, what we manage and specify T. And then, you know, Hari is here, who can walk us through the play by play of how the migration is happening or has happened in Cisco, right? So, you know, as everyone knows, Cisco is here, because, you know, company, we have close to 1,500 Oracle databases that we manage within ourselves, right? And out of which, close to 500, 350 are production databases. These databases actually, if you look at it, we develop in-house applications which are supported on the plain vanilla Oracle databases. Or we have 10 to 15 Oracle installations, Oracle ERP installations. You can name as, you know, services, contracts, you know, finance or, you know, supply chain and things like that. And not only that, we also have a lot of middle-grade installations, IDM, SSO, agile, SOA, web logic platform and everything, right? So that's where the complexity comes into the picture. And we also wanted to make sure that we operate in a burn class so that the TCO is reduced on a year-by-year basis. And now with that, we had a huge opportunity for us to migrate into UCS to reduce the cost for us. And, you know, Hari has been the lead architect who actually helped us in migrating from the legacy servers to UCS. So with that, Hari can give us back to us. So before we get there, though, how long did it take to migrate? It's a very good question. Actually, it's not that hard. But before I answer the questions, right, I have to divide my complex, my environments into, you know, three tiers maybe. One is the small-sized environments and medium or mid-sized environments and the highly complex environments. To answer your question, right, it depends on what type of environment that we are talking about. Alright, so we'll hold that. We'll go to some of the architecture. So this is the fun part. Okay, so let's get into the architecture. Lay it out for me. What was the challenge? Now the decision gets made. Everything rolls right to you. Go make it happen. Do more with less. Here's all the UCS. So just to add a bit to what Ravi mentioned. So the first challenge was to how to mitigate some of the current challenges we had. That's what prompted us to migrate. That was not the only part that Ravi covered, the TCO part also. There was just one more point, which was the main one, is how to transform our IT to adopt new changes, you know, the new trends in technology, like big cloud, virtualization, big data, a lot of those things. How do we transform ourselves to be ready for such new trends in the industry? So those were some of the things on our plate that we had. No guiding principles. Basically, you got to check those boxes. Exactly. Absolutely. Now, what did it take to migrate? So again, as Ravi mentioned again, you know, you have the simple environments, you have complex environments. Any migration typically depends on various factors, like size of your environment, the complexity, the criticality, business criticality. So these are some of the things that you need to consider before you go to any migration. But there are environments which could be very small in size and which are very critical. So it could be a matter of just a few hours maybe. But there could be environments which are so huge, like in one of, we'll talk about one of our use cases. The database is like 40 terabytes running on three huge monolithic servers. How do you move that over to UCS? And why would you want to do it? So we need to go into what were the challenges with those. So there were challenges for plenty. And our scalability was biggest, right? And so we had this, so this particular environment we talked about is like 40 terabytes in size now spread across three rack clusters. We had been constantly beefing up the servers to some extent vertically. But at some point we hit some technology, some technical limits where we couldn't scale up, scale it up any further vertically. So only option was now you add fourth cluster, fourth node to the cluster, which is huge, expensive. So these were some of the technical challenges, the cost of operating this. So all of those paid a factor in it. So we decided, you know what, because for all those guiding principles that we talked about, UCS was the best fit. It let us move from these huge monolithic servers into much more horizontally scalable servers, smaller servers, which gave us a lot of benefits in terms of performance, scalability, reducing our cost, improving our quality of service. So all of these we achieved by moving it. Now I could go into how we did it. Now so was it just gear swap out? Or was it like the service profile capability? What were some of the things that were like, give us the highlights. Sure. So what it involved in our case was, in this case, the 40 terabyte database I'm talking about. So these were all like some legacy systems, like HPE systems. So we had to move that over to, moving to UCS means we had to move to Linux. So the biggest challenge was converting them to the Linux format. Indianness, from the bigger Indian to a smaller Indian. So platform conversion. So that was the biggest part. And size of course posed a big challenge. So we adopted different best practices, the best of the technologies to do that. So when it came to Oracle, we used technologies like R-Man conversion, which would let you do Indian conversion now. And then when we did this, we also had to keep a note on how we actually spread these environments out. Like the services, you club the applications in a way that you measure their IO requirements. You group them logically based on their throughput requirements, and there's their functionality. So all those analysis went into it. We grouped this application across the 12 nodes that we decided. And why 12 nodes? I'll come to that because we wanted to maintain the same compute capacity. Because when we started, we did not have a good comparison on how UCS would scale. Would it be better? Would it be worse? So the safest approach we took was we'll go one is to one. We said, okay, we had X number of CPUs here. We'll go with X number of CPUs on UCS. So we decided to go with 12 nodes. So that was the methodology we took. So then a whole lot of analysis went to it. Now to your point about the service profiles. So service profiles did play a role in the UCS provisioning. The way you provision UCS with service profiles makes it so much easier to provision hosts. And even in terms of resiliency, say, a new host goes down to swap it with another host and apply the profiles over there and get the business quickly up and running. So all those things played a role. So software is really critical? Exactly. So John, I will give you an example of how fast the UCS provisioning has helped us while I was at the middle of an incident. Basically, I had to handle large volume of the transactions hitting in one of the ERP systems. And we had four middleware tier servers where we had to add eight more servers. All of these need to happen in the middle of incident with you in one hour. This is where the service profile came into picture, saying that using the same serving profile can quickly provision the UCS servers for me to add the Oracle mid-tier services on top of that. And that really helps us and so on. So yeah, definitely some of the UCS features like the management is a key asset for those who are doing the UCS management today. So the UCS had those features on the management side, the provisioning side. So the big management, all that stuff, all kind of plug and play. Exactly. That's where we are able to say UCS is the perfect platform for the cloud today. I mean, if you look at Oracle today, right? They announced the cloud platform, but not just Oracle, right? If you look at, you know, a lot of companies today in the world, they are moving towards Oracle. And customers also wanted to host their platforms or services in the cloud. And for those, all of them, the management, the access to the servers and the control and the ease of provisioning with the tools that are provided at the hardware level, with the service, right? It basically provides ease of management with the UCS. So that's the key. So I guess the question is, what lessons learned did you get out of migration? So there are a couple of best practices we followed, or we learned as we went as we did the migrations, and there were a couple of good lessons we learned. And the key lesson is like, you need to know your application profile, the characteristics. That is the key thing. That's more of a good practice, best practice. Now, what are the lessons not, you know, things like patch analysis? When you move from a platform to another platform, you need to work with the vendors, you know, be it Cisco in terms of hardware, be it Oracle, when you move from, say, a legacy system to a UILX. So patch analysis plays a very critical role. I mean, the more lessons it's going to be, because we practically face some challenges in that. So we ran into some corruption issues, for example. So we knew there was a moan bug in Oracle. Unfortunately, during our testing, the patch was not available. We kind of underplayed, we kind of didn't think that patch, that issue to be so critical. Unfortunately, right after we went live, that bug did hit us, and we ran into a corruption issue. So, of course, we got the patch and we rectified it. But that was, I think, one of the key learning for us, you know, do your patch analysis thoroughly, know your applications. That's the big data angle. That's the big data angle, having all the data information. Absolutely. So that's a key thing. So you want to add something? You're good, yeah. So that's pretty much it. I would say that the main application behavior, the profiling, the IO requirements, analysis of everything. That's a very reactive situation that the data helps you on. So to say there's a heart bleed or, you know, another breach that we just had last week in these systems, you want to know fast. Absolutely. And which systems are vulnerable? So you want to know fast, as well as you should be able to react fast to it. And that's what I think UCS plays a big role. So you have the central management console, right, management server framework, which lets you, say, if you have to patch an environment for security vulnerability for a heart bleed like scenario, you could just push it through the central management server to all the nodes within the management framework. So it's so much more faster and easier to manage with so much more less downtime. I think that's the biggest benefit that UCS has. Yeah. And some of the lessons learned were that, you know, when we are basically migrating such a large systems, the systems that we migrated was actually like a 32 terabyte system. We were with the three monolithic servers, 128 CPUs each. So when we are doing that, you know, how do you lay out the architecture of the new systems? Our system was one of the largest ERP implementations when we did the work with the 12 node Oracle Rack. And, you know, if you look at it, right, one of the challenges that we always face is basically how Oracle can transfer the data across the servers. That was the biggest challenge. And we had to go through multiple incidents in the past. But after we basically, you know, perfectly architected the, you know, systems with the 12 node Rack, and we kind of did the applications partitioning as well. With that, we were able to reduce a lot of the transfer of the data across the nodes, which basically reduced the incidents for us. Okay. So, John, you touched upon the migration part. I just want to add, since we're talking about 30 to 40 terabytes, it took us about 24 hours to migrate this database. I think that's a commendable, that's a huge achievement for us. That's good. That's very good. So, let's bring this back to the customers out there that are watching. You guys were thrown into it. I'm sure the phone call came down from high, hey, get us RIT, UCS, because you're in Cisco. But you're the internal customer. So, you're still going to be hard on Cisco. You're not going to just roll over. Now, you may have to, you know, get the tech geeks involved, but you're the IT guys. You need to make sure it works. Because at the end of the day, the employees don't care what the boss thinks. They're going to want to have a functional system. So, other IT groups out there, what do you say to those guys about UCS? Why should they move with Oracle? Why now? Share your opinion. Yeah, basically, I think, you know, the existing challenges are wanting, right? You know, scalability is a bigger problem with legacy-based servers. Performance of applications is another problem and dealing with the day-to-day problems. So, to me, they are all, you know, ongoing existing issues and things like that. But even when you look at the futuristic solutions needed as well, for example, big data, you know, today, when you move, you are company yourself into big data adoption. Or when you are getting a lot of, you know, data from your devices that either your employees use as a contractor based using it, you know, today, big data applications pushing the server to its limit. And if you wanted a platform that can scale better, UCS, you know, can help you better, right? Virtualization. UCS actually offering, we talked about the management solutions, one of the best in the industry. Plus, it also offers a capability to manage the virtual servers itself. The VMs were, you know, switching, right? So, it offers a capability to manage both physical and virtual. So, we talked about the cloud in the past as well, right? So, if somebody is looking for a futuristic platform which will, you know, cater the needs of big data, cloud platform, and virtualization, then UCS is the answer for itself. I think to summarize, maybe it's like the cost benefit, the performance, the scalability, and, you know, adopting the new trend to be up to date with the new trend, I think UCS is definitely a perfect pitch and that's a way to go. So, Harri, talk about the guys out there, but the naysayers, I'm too busy, I'll get to it later. There's a lot of people out there who might not think, maybe I shouldn't, maybe I should wait. What do you say to those guys? I mean, we're probably in the same boat, initially as IT guys, right? But once we went into the cloud, when we started the journey, I mean, I think it's better done sooner than later. There is no point postponing it. If you could realize the benefit right away, why wait? I mean, I think the benefits are very fast. People realize the benefits really quick and it's almost instantaneous, I would say. So, I think, actually no point waiting, you should probably jump into it. One thing that I want to add is basically the benefits we saw, right? The performance benefit, both for online as well as the batch applications, we saw anywhere from 20 to 30% increase. And cost benefit, the operating cost, we can definitely say for sure, we reduced by 60%. And the floor space, I think that's the biggest challenge in today's data center, right? Yeah, huge. Exactly. When we have been building the next generation data center in Princess and in Allen, this actually helped us a lot in terms of reducing the, you know, space. And not only that, power. Actually, we were able to, you know, reduce the cost of the power by 300%. I think here are some inherent benefits of moving to UCS and importantly, right? Your customer satisfaction. Imagine you have your legacy monolithic servers where it frequently goes down. We are not even able to take the orders. We are not even able to fill the way for the customers. Now, when we migrate your applications to UCS with Oracle on a rack, you have the scalability, both vertical and horizontal. That provides a immense customer satisfaction. And that's exactly what we saw. And that's exactly anyone will benefit when they migrate to Oracle on UCS. I think the footprint alone, justification, could justify a technology refresh. I think that's kind of like in the refresh cycles where you're at. Guys, thanks for coming on theCUBE. I really appreciate it. We really enjoyed the practitioner perspective, although a bit biased with Cisco. We're in the Cisco booth, of course. But, you know, I know from other companies like Cisco that have IT departments, they don't roll over. They still have strict requirements, but having that perspective is really good insight. Because you guys have a big Oracle shop and it's a big task. Thanks so much. Congratulations on your success. Yeah, thank you very much, John. I really appreciate it. But I do want to close the thought by saying that there are some white papers out there. We have completed the benchmark and everything with what we have did for our customers. So they can actually look at the internet as well. Awesome. And where are those benchmarks out on the site? Cisco site? It's in the cloud as well. It's in the Google as well. I think if anybody has a question. Cisco documentation, white papers section should have it in Cisco site. The TPC benchmarks are all in there. You guys have some good benchmarks. Again, they're just throwing the benchmarks around. But that's good for people to get the benchmarks, to just get a taste of the kind of what's going on. I don't even know if those benchmarks we do mention how we did the migration. Some of the steps we thought were solid. You guys are very transparent. Yeah, very much so. Well, congratulations. We're here inside the cube. Inside the Cisco booth is the cube. I'm John Furrier. We're back with our next guest after this short break.