 Live from Boston, Massachusetts, extracting the signal from the noise, it's theCUBE, covering Red Hat Summit 2015. Brought to you by Red Hat. Now your hosts, Dave Vellante and Stu Miniman. Welcome back to Boston, everybody. This is theCUBE, we're here live at Red Hat Summit. SiliconANGLE Wikibon, we go out to the events, we extract the signal from the noise, we get a great guest on Dietmar Fauser, Vice President of Architecture and Quality and Governance at Amadeus. Dietmar, thanks very much for coming in theCUBE. Hello. So tell us a little bit about Amadeus, for those in our audience that aren't familiar with the project and the organization. So Amadeus, we are a company providing IT solutions to the travel industry, so we are one of the biggest player in this industry. Traditionally, Amadeus is what we call a distribution system, so we link what we call providers and subscribers, so travel agencies, online travel agencies, TMCs, so travel management companies like AmEx, with the providers that are essentially airlines, hotel chains, car rental companies, cruise lines, whoever has a product to sell. So it's really kind of an electronic marketplace that matches the demand and the offering in the industry. And it was born out of a pan EU effort, right? Is that correct? Yeah, Amadeus has, it's sort of born on the internet, just a little bit before, but then pivoted to the internet, right? Kind of, so Amadeus has been created by three big European airlines as a European project. That's why we are structured across Europe, so headquartered in Spain, development essentially in France, and the operational center is nearby Munich in Erdinger. So there was a political intention to do this. So it was one of the big kind of electronic marketplaces way before the internet existed. So we offered, at the time being these green screens, terminal access, but at a very large scale. So Amadeus had quite quickly 350,000 terminals connected to it, and then in the late 90s came the internet, big explosion of course. So kind of a Sabre competitor, is that right? Absolutely, we are the other player. Right, the two big whales in the business, then of course, all kinds of, disruption going on in that business that you have to respond to, which kind of leads me to, so you have all these competitors trying to come at you, using disruptive technologies. How have you responded? You've got a development center in the south of France. I'm sure DevOps is fundamental to your culture. You're here at Red Hat Summit, open source. We're talking Paz all week. Talk about the culture, your development culture. Okay, so we are, I would say, a pretty aggressive adopter of new technology. So we were founded as a mainframe shop, running on IBM, TPF technology. Late in the 90s, we switched to open systems, like Unix and then Linux systems, so highly distributed systems. Well before, kind of open source middleware solutions were available, and now we see the next step of technology changes coming in, like big cloud management solutions, hybrid clouds and distributed environments, and we want to ensure that our materials is ready to take these opportunities should our clients ask us for this, and of course bring it into our own data center because we expect a lot of advantages in terms of automation and less human intervention, so less error-prone operations. So yeah, so you've seen quite the evolution, I'd say, from TPF to what used to be called open systems, it was the Unix era, and so now you've got a culture of DevOps emerging in the world, where are you guys in that whole journey with regard to sort of that DevOps culture? So I think like most significant are almost all companies, I think in the IT field, we look at this very actively, so we have a traditional split of development operations, so currently there is a kind of a handover from R&D to operations when we ship releases on new products, and we clearly believe that we should overcome this and tie the organizations more together, so in recent projects, we implemented DevOps teams, so joint teams with people coming from both operations and R&D organizations, and we see a clear benefits of doing this, so there is a better flow of information and I think the tendency is like across the whole industry is to go much more in this direction. So let's paint a picture of your organization, so you've got existing applications that have been around for a while, you've got sort of new apps that I'm sure you're writing for the web, and then you've got others that maybe you might want to move to the cloud from on premises, but so you still have mainframe apps? We still have a little bit of mainframe around, so the firm sunset date is end of next year, so end of 2016. And off the mainframe. Of course, so we track this on an almost daily basis, the CPU usage going down according to expectation. A very long and complex project, it took us years and years to get this, and now it's coming the next step, the containerization and making applications ready for cloud operations, whether it's in our own data center or perhaps in the future in other hosting environments, it's the farthest disruptive step for us because our applications are reasonably modern and sticking them into a container and having them scheduled over a container management system like OpenShift with Kubernetes. It's not such a disruptive move for us. And so your primary sort of infrastructure is what, Linux, obviously, and some Windows or? Yeah. Right, like everybody else? So essentially Linux, from a pure capacity point of view, it's way beyond 80%, surely in the 90% it's running on Linux. Our e-commerce environments is running on Microsoft Java and we have in mind over time to move this onto a Linux standard computing environment, the same that we currently work on. But we know that this will take some time and effort. The good thing is with containers, you don't care so much whether there is Java running inside or Python or C++, like most of our applications or whatever actually. So, Dietmar, it's pretty early in the whole Docker discussion. Can you help explain to your peers what led you down the containerization path and what you've benefited from it? Yeah, sure, I mean, Docker is, I think, I believe the fastest adoption that we have seen in the whole industry ever. So many of our teams naturally look towards new technologies with very high flying and curious engineers. And so there is a natural adoption. So the benefits we expect from Docker is the standard operating model, like scheduling standard containers no matter what they are doing over the infrastructure, having a more traceable deployment chain because the good thing with the containers is that you're entirely sure that the same thing, it's signed and trusted containers, the same thing runs on test systems and on production. So there is a software engineering advantage and there is a runtime operational advantage to have standard containers. So really kind of the life cycle management of your applications? Yeah, life cycle management and operations, both are equally important. I think from an agility point of view, life cycle advantages are important. From an operational point of view, it's important that the system is uniformly run so that there are no kind of idiosyncrasies depending on the technology you use because this ultimately leads to error prone procedures or erratic behavior when there are issues because people have to see, okay, this is the e-commerce environment, I have to behave differently than in the open systems. So your team has a lot of Linux background, did that help with the adoption of the containers? Oh yes, we have a very strong Linux culture. It became pretty evident quickly that container would be something that flies. From a software point of view, it's very close to what we do with software anyway because containers are layered, you have a version on it and you build a container almost like you build a software release, so it's from an IT point of view, quite a natural adoption. I would say our strong Linux culture helped us more on the OpenShift Kubernetes adoption than on the Docker adoption itself. We dig deeper into the details when we talk about the OpenShift platform really. So you've also got governance and some people are concerned about security when they talk about containers. How did you sell this up to your management chain and any kind of challenges or lessons learned that you could share along those lines? Well, we systematically look into security. So we don't believe that containers would add a significant risk. I mean, the whole thing is that you have to have a trusted source from where containers are running. So we have what we call a DML, a media library. It's a thing where containers get stored where the runtime systems pick it up and running. So what we have to be sure that only trusted development teams can commit a container into this environment. So we will not run containers coming from Docker Hub and right on our production system, of course not. So we don't intend to open Amadeus as a standard a path where everybody can run its stuff, of course. So maybe talk a little bit more about the Amadeus cloud, the platform, what's the objective? How do you, are you building an ecosystem around that? Maybe talk about that a little bit. So the principle objective is simplification and automation of our operational environment. So as I said, we have Microsoft environments, we have Linux environments and the Linux environment is very big. There are little deviations here and there. So what we call the Amadeus cloud services is really rolling a standard scheduling and operation management system into our own data center. Our kind of definition for what a cloud system really is, because I believe there are many of them out there, is that with these systems we stretch the resource management to the size of a data center, which means you have global resources available and you let a computer system choose which resource to use for a given task. And currently we do this in a very stringent, but much more, I won't say old fashioned, but more classical way by saying, okay, these machines are attached to this application. It's not fully automated at this point. It's not fully automated. It's very much automated, but it's not... But it could be, but you've chosen not to... The machine's not making the decision as to where to place the workload. Is that right? Well, the difference is that currently machines kind of are fixed in a fixed way assigned to applications. So a machine is named according to the application it will run on big cloud systems. This doesn't make sense anymore. You just see a machine as an anonymous piece of... It's invisible. Of capacity and it's invisible and it should not be up to a human to choose which is the right piece of capacity to run a given task at a given moment. It's, if you think this through, that's exactly what Linux is doing. I mean, you have cores, you have memory available and you have many processes running on the operating system and the operating system chooses to which task give a given core or given CPU, which memory consumptions you get, how much access to computing resources you get and what we see with Kubernetes and these environments open shift is that this model gets stretched to clusters of machines like tens of thousands of machines in some extreme cases. So we kind of move the scheduling paradigm up to the whole data center. The data center is the computer. That's what we try to teach internally. So there is a way to go, of course, to implement this. Right, but it's the vision. It's the vision. How about your database technology? What's, I mean, a variety of different databases that I can imagine? Yeah, you're like an oracle shop. Like in big environments like us. Primarily Oracle, we, as I said, we adopt technology fast and early. So we are very, very huge, no SQL shop. Even with an Oracle, we use key value data models. Just oracles, key value store. Well, we have built this our own. It's very simple to implement a key value store on Oracle. You have a table with two columns. One is the key. The other is just a binary object. But you use that platform too. We did use this this way because no SQL solutions didn't exist back in 2000. And now we use other technologies like many of us. We use MongoDB. We use Cache-based, Memcache-D a lot. So if Oracle comes to you and says, hey, we got this great new Paz announcement, new cloud. You don't have to go anywhere else. Red stack, extend Oracle into the application domain. Forget that red hat stuff. That's just fragile. What's your response? I would be... You throw them out? No, no, no. But I would surely argue a lot with them. I mean, we have a lot of trust in what we are doing with Red Hat. It's not a tiny startup that comes along. They have proven that they can bring enterprise-ready things to the market, like Linux. Everybody is using their Linux distribution, so do we. We are a strong believer in open source. So this is one of the key arguments, the fact that we profit and we contribute to something that is of common use. The infrastructure is not something that brings us a competitive advantage, really. It's our applications, our technical and functional know-how to build solutions for the travel industry. So we quite like the idea that we share our collective intelligence to build the 90% of the computing stock that is not the application that everybody needs at a reasonable cost and too high quality. So let me play, I hate being in the position of playing Oracle, but you said I would debate them. No, no, no. Okay, so they're going to come in and say, the advantage, I hear a lot about integration. The advantage they're going to say that we have is this full stack integration from silicon all the way up through the application, and that's going to give you better performance, better reliability, better recovery, better everything, better scale, blah, blah, blah. The open source can't match that, is what they're going to say. I'm hearing, you don't buy that. Why don't you buy that? No, I don't, for a couple of reasons. First, the biggest cloud environments are built on open technology. So Google, Compute Engine, AWS, there is a lot of open source technology in there. Secondly, there are indications that in the future some of our clients might ask us to run the solutions not in our data center, but closer to a given booking source, for example. So like a lot of other companies, there might be a need to distribute the computing and not to master the infrastructure over which you run it ultimately. It could be any provider, a client of us, or whomever. And it's really important to decouple the applications from the underlying infrastructure. And this is why we build our applications around the OpenShift platform that can run over a variety of infrastructures, so OpenStuck, or VM there, or whatever solutions we choose, or bare metal. It's important to decouple the application from the infrastructure to give you more flexibility, freedom, choice, speed. Yeah, not necessarily speed, I would say, but choice, freedom to operate. So that's a key point. So the examples you used. Imagine for one second you are rhetorical and you face the need to run something on Google Compute Engine. You would have a, it's possible, of course, but it would not be the most important one. It's all Java. I'm teasing. It's not all Java. I'm teasing, that's what I'm saying. That's what Larry Ellison would say, it's all RINJA. I'm pushing because I'm trying to understand that's the only reason, because what you talked about, Google Compute Engine, Amazon, et cetera, those are built on open source, but they're highly homogeneous environments. And Oracle's coming at you with a highly homogeneous environment. My question is, are you able to replicate, or do you even need to replicate that homogeneity in order to replicate the capabilities of, let's say, those public cloud providers? As I said, homogeneity, especially for the infrastructure parts might be not the right parameter anyway to take into account. I mean, there is a given set of heterogeneity. Once you say you want to make a platform ready to run it on different infrastructures, homogeneity is a bit gone, huh? Yeah, that's great. So Dave, I got a question from the crowd actually. Deepmar, they said, I mean, you guys are moving pretty fast. And the question they have is your customers, the people that run on your platform, have their operations on Amadeus. You know, bring us inside that communication because we got to expect that there's times that they actually aren't ready for it or holding back or how do you bring them along? How does that give and take of customers that are ready to move fast and those that move a little slower? Well, there are two aspects of the answer to this. First, I believe historically our customers have a lot of trust in what we are doing. I mean, we operate for them, especially for the airlines. We are a very big provider of LNIT systems so we operate flight inventories, airport operations for the airlines. And our customers are used by the SAS models of cloud services service to outsource this type of stuff to us and they acknowledge that we moved from mainframes to Unix and they will also acknowledge that we moved towards cloud-based systems. There is another thing, we don't explore this actively but a lot of our customers are in discussions with us to have other third-party companies extend our own capabilities, like serviceability of our own solutions. And the more you move towards standard containerization, the easier it would be in the future to take third-party software to extend our own core capabilities on our platform. Should we decide to do this? As I said, we don't intend to position our materials as a pass to do this but it's something we keep in mind and this is also why we believe that containerization and standard scheduling is so important. So every vendor has an open-source strategy. Put that in quotes. Oracle does, VMware, obviously Red Hat lives, breathes, eats, open-source, certainly IBM, Microsoft. Microsoft is opening up. Definitely opening up. As a buyer, how do you evaluate, first of all, I'm making the assumption open-source is very important to you. So that's a key criterion for your vendors to have. How do you evaluate, because they all come at you with marketing. Oh yes, we do open-stack and we love Kubernetes and Docker, blah, blah, blah. How do you evaluate the validity, validate their story? That's a good question really because I think there's no single scientific answer to this. So it's really a mix of trust that we have in a given company. In some situations, the funding a company has, if it's a smaller company and not like Red Hat, the adoption of the industry. So Kubernetes, for example, I mean, there's Google behind Red Hat, quite a bunch of other big important companies. So it looks like something that will be around for a significant time. For sure, the fact that it's fully open-source is very important because it gives us the freedom to put our fingers in and if need be, maintain it ourselves. We don't want to do this, but the possibility is there. There is an aspect of forward-thinking, so the modernity of the solution. So it's a mix of given parameters, but we work with a lot of the companies you were setting. We are a significant customer of VMware, so we talk to them also and we have had extensive discussions with them around Cloud Foundry. We use their open-stack implementation, for example. So it's a mix. We try not to put all of our eggs in the same basket either. And many of those stories are compelling. I mean, these are smart people. They listen to their customers. They know that if they don't listen to their customers, they're going to end up like Digital Son or Data General. Absolutely, there are great companies, intelligent people, and they see the threat of open-source solutions coming around and they adapt to this, of course. But at the same time, they know that if they adopt publicly Kubernetes and Docker and things like that, that they can freeze the market on the upstarts and they can maybe take their time. Is that changing though? Are the market pressures so great that companies like VMware, certainly IBM has adopted its versions of open-source and does very well. The HP now is very much on the bandwagon. Are they going to be pressured to really act, in your opinion, as opposed to sort of fud the market? I'm convinced, yes. Yeah. Times have changed. I believe that there is significant pressure. As I said, I really believe that the essential parts of the whole computing stack, so we have seen this with Linux. We see this with cluster management systems. We see this with middleware solutions. The whole stack really is available through open-source solutions, which is a great and a massive game changer, isn't it? And so Red Hat has this brilliant model. I mean, it's so counterintuitive. Give, spend like crazy, invest like crazy, commit, give it away for free, and then support it. And then in retrospect, wow, that's genius. But in other companies have tried, many have sold. Red Hat has bought some, VMware has bought some, I give others. Why do you think we don't see more success with that type of model? Is it just too early right now? Is it too hard? Yeah, it's not an easy thing because you need to have trust out there so that people really believe in your future, that you will be around. I mean, for us as an enterprise, it's really important that a partner stays around and we value this. We like the idea of supporting companies like Red Hat, obviously financially also, through taking their support services. And I believe that there are a lot of great success stories out there also. I mean, the Mongo guys, for example, are doing great. A few other companies are doing great. I think the Docker IPO one day will be a healthy day for the guys who have found it. I mean, Hortonworks has given a great bid for it. You know, their initial IPO or the income statement wasn't that pretty, but starting to show some potential there. But still very early, it's all less than $100 million. Yeah, yeah, and for example, you were questioning about the database environments. There are really a big number of players out there and every day new ones come in, Cockroach, DB or all that. Yeah, DeSandra, Aerospike, Open Source now. You name them. And so here it's sometimes not evident to really see what consolidation and who will survive on this game. Yeah, so it is very early. It's very early. We all adopt these type of technologies, but the end game is not yet known. Yeah, so talk a little bit about your data strategy. Are you doing Hadoop, for example? I mean, you're using a lot of the key values to do Hadoop. Yeah, I mean, historically, we are more transactional shops, so we have very, very high transactional volumes, but we move more and more into analytics, including real-time streaming type of analytics to better understand how our system is used to be more sophisticated when it comes to fraud detection and quite a lot of other initiatives ongoing. So we use, obviously, Hadoop. The last time I checked, Amadeus was hosting something like 30 petabytes of data, so there's a lot of data around. And the quality in your title is part of that quality data quality? Are you the de facto chief data officer or no? Not really, no, no. I'm more... Software quality. Yeah, I'm more there to furnish the base technical capabilities. I'm not overseeing the data scientists or trying to harmonize the data structures in the company itself. What about the governance, basically, in your title? It's pretending to R&D, so parts of my activity is to give what we call, historically called development supporters of engineering services, so we provide a standard development environment. We ensure that gradually the applications go towards containerization. We have an internal technical policy, so we have a catalog of what we call non-functional requirements like security requirements, scalability, recoverability requirements, and we try to ensure that the different various application teams adhere to common practice, which is a challenge in itself in a big distributed job. What's the most fun part of your job? What do you enjoy the most? Me personally, I'm really very much a new technology. I'm a distributed systems specialist, so I quite enjoy what's happening currently with Kubernetes, and I'm looking a lot for the in-memory data management system, so that's where I get possibly the most technical satisfaction out of it. Well, it's your renaissance of distributed systems now, for years distributed systems have been so limited, and many just didn't work at scale, and web scale has changed that, so very exciting times. David, thanks very much for coming on theCUBE. We have to leave it there, but it's really a pleasure talking to you. Thank you. All right, keep it right there. Everybody, Stu and I will be back to wrap up day one at Red Hat Summit. This is theCUBE. We're live from Boston. We'll be right back.