 We may as well make a start. The rest of our panel may walk through the door in a moment. If they don't, we have more time. That's good. My name is Paul Miller. I'm going to be moderating this conversation today where we're going to learn a little about real enterprise experience of really using OpenStack to do production workloads. One of the things we're keen to do is to get audience participation, to get all of you involved. If you have questions or comments as we speak, please do contribute them and we'll get a conversation going and hopefully learn from one another. The other thing I have to say before we start is that Alison at the back, taking photographs, has said that everyone who is here can have a t-shirt at the end to reward you for staying and for contributing and for sharing your perspectives. If you see Alison at the end, she'll give you a voucher which lets you pick up a t-shirt from the Intel stand. So let's get going. As I said, my name is Paul Miller. I'm going to be moderating the conversation. I have an esteemed and rather small at the moment panel which is going to get bigger in just a moment, I promise. Yes. So I'll let our existing panelists introduce themselves. First off, Carmen. Good afternoon. My name is Carmen Rimi. I'm a director of cloud engineering over at Workday. Workday is a successful software as a service company focusing on HR and financials and other products over time, including big data. Good afternoon, everybody. My name is Patrick de Busescaire. I'm working for Numergie. Numergie is a public cloud company. We started two years ago. We are fully engaged with OpenStack and I'm going to tell you what are on the road. Nicely, we said journey, but the journey is going to have some rapid. Rapid is something we need to mitigate. So that's the return on experience we're going to share with you. And I hope you're going to have a good time. Thank you. So sticking with you, Patrick, what are you using OpenStack itself to do? So first of all, I mean, when we started two years ago to build a company, we were thinking about the visions. And when in terms of visions, we were defining should we have vandalock or do we have to think different way. And when you said different way, what kind of decisions regarding the risk, we're going to be able to assume. So we started to think about OpenStack. It was the early days of OpenStack, but there is some very true experience coming from Rackspace and others. So it was inspiring me somewhere and the team and said, OK, if we go to OpenStack, which release are we going to start? So we were willing, I mean, Grizzly first, and we're starting with Grizzly right now. We are I South, very close to move to Juno. So Juno will be probably on the early stage of next year. And I have to tell you that we got the breadth of the foundations. It means a lot of innovations coming from the innovations from everywhere. And we see it also during the first year where we decide to go with OpenStack to see the big player coming in a game. So big player, you know them because they give you a lot of t-shirts, caps and others. But the big players, it's another things to think about. I mean, is it solid? Yes or no? Can we think about, I mean, an engine would deliver services to company and those companies could be stock exchange. It could be very young company, which are looking for services and paying the services only when they use it. And we think that OpenStack was the right way, I mean, to go with, to think about the new infrastructure service. And our platform is commercially available since last June. It took us a year and a half, I mean, to design, to prove of concept. And finally, we were going on the road with five partners, which has been to criticize our jobs, the design, and they get the confidence in numergy and the way that we have implemented OpenStack. So I will tell you how, when and what was the big issues we were fixing and we have fixed. So, just sticking with you for a moment, now you were essentially building a new company, trusting OpenStack. Now, OpenStack grisly wasn't the most dependable piece of software to launch a new company on. So that was quite a leap of faith. Yeah, you say in English, mitigate. I like this word because we don't have the equivalent in French. So mitigate means that you are taking risk and you are looking for the value. I would say we find the value, but also we find some other things than company you are dealing with for many years should I have to buy or should I have to make. And that's a big, big, big concern as a CTO when you're borrowing to the team and talking to the shareholders, then I have to have their confidence. They know that they are looking for cost, means that license or software is not an issue, it's an option. And we were, let's say, commit to say, with good engineers, with good schools, engineer schools, you can get the best of breed, but also you have to think about release management because this market, it's not only skills, but as soon as you develop the skills, competitors are looking for you. Not only to use the platforms, but use the skills you develop with. So the risk, it's financial, it's about skills, and it's about partners, good partners to deal with. And for France, OpenStack three years ago was an interesting journey again because only one company was obviously satisfied to talk about OpenStack and the rest of the market was much more, I mean, two-PC talking with OpenStack. So when you said, I'm going to sell to EDF as we did with a good partner and I'm going to tell probably some words about HP because we were working with them and designed the CloudOS on the early stage and the French company, as numergy is, bring the inspiration to a very large company and said, okay, if you want to operate this way, you're going to go directly to the world. So it was the right combinations between big player, small player, bring the energy, bring the visions and finally how we want to operate for customers infrastructure as a service as a public cloud. So that was the big questions we fixed. I think it solved for the time being, but the next release and after release, it's always another game and that's where the topic is about release management because release management from OpenStack is quite an interesting one but when you talk to very large company, I mean they are doing release management every six years and we get every six months. So how do we fill the gap? Yeah, and that's the topic we want to dig into a little in a moment. Workday, what are you doing with OpenStack at the moment? At Workday with OpenStack, we've had integration paths for quite some time. We allow our customers to submit workloads, Java workloads that we run in our cloud and that was one of the very first workloads that we've used to migrate over to OpenStack, leveraging the existing sort of virtualization technology that's already there. We're also looking at the whole breadth of our workloads at Workday and in the fullness of time the hope would be that we use OpenStack to run the entire data center as a combination of ironic, a combination of KVM and then also containers. So each of those models were and when appropriate. So I mean for both of you, you were taking a risk of one form or another in moving to OpenStack. What was the evaluation process you went through in comparing it to some of the other ways you could have solved the same problem? Start with you. Sure. So at Workday, we are a couple of different things actually. We started even with evaluating operating systems, Ubuntu versus CentOS. We looked at different network options, different storage options and each of these major pieces within OpenStack, even Compute, we're looking at containers versus KVM. For us, the two, maybe three fundamental things are, does it work? Which isn't always the case. Does it perform and how secure is it? Obviously you have engineering teams, smart people, dive into those issues and tear it apart, put it back together and come to a decision, converge on what works best and then move forward with that technology. That's kind of what we did at Workday. Yeah. As a public cloud service provider, I mean the two big issue was should we have to virtualize that's okay. Everything is agreed on that. But what kind of virtualizations? And as Kevin said, what kind of kernel I should deal with? There is a lot of options there. So you can decide to build your team, which are contributing to the Linux foundations. That's one option. The other one, it's looking for what is available on the market. So you should talk to your engineer teams and they have multiple ways and multiple feelings to say that's a good one. That's a really, really, really best one. And at the end, when you are looking for, you think about what kind of supports I need. So that's exactly what Kevin said in a very political way. Nice. The other things where we were very, very sensitive was the networks because if you say that my revenue coming from the public cloud, you should have a network. With no networks, you have a very nice platform but nobody cares about it because nobody can use it. We were adding another point at Numergym, because we love things which are complex but under control. We are saying to partners, we don't talk directly to customers. Of course we are going visiting customers with the partners, but we don't sell directly. We don't have any shops on the internet only for the very small and limited VM. So is it open stack or not? It's a bad cost. So choosing QVM and which QVM is the best one. And when you think of the best one, is it my choice or the company which I'm targeting as a customer will tell me what is the right choice. So bundle the kernel itself and the network piece because probably everybody knows then the first network model belong to the open stack. Folson, Grisly. It was unpredictable in terms of stability and I'm very, let's say, clever when I say that to you. Everybody are thinking about the network piece. It's extremely important for a cloud service provider as we are giving the network access its key because it's about revenue for us. It's about also customer experience and it's about flexibility. So we were very committed to say network will be happen at Numergy only if it's virtualized. So we were picking up on the early day SDN. And we have implemented SDN with the weakest piece of Neutron or Cantrum depends on your failures. And networks has been really implemented by Numergy so we are using SDN from Nuage, Nuage Network because we were directly linked with the labs. We're getting the very, very late fresh driver and we were operating also a canonical kernel. So those three elements bring support. Even to support it's quite difficult to explain when you get an issue because everybody is looking for what I did there, what kind of documentation is available, et cetera, et cetera, because you are running fast. You should focus on the business and also you should focus on what should be controlled. And that's the numbers of conversations I got with my colleague is there. It's about productions because it's nice to design to the driver but at the end it should be operational control. So that's the main piece we were focused before to say, OK, OpenStack is the wire road for us because we believe that with OpenStack we can run multiple instances, we can have multiple hypervisors and in Numergy we said we want to produce the digital energy for today and tomorrow. It means that the engine has to be very solid, very reliable and significant also in terms of saving your cost for customers. Did you look at things like Open Nebula, which was perhaps bigger for the European part of the audience, they may be more familiar with it than the North Americans? Well, I will talk only for my name. I've been working 25 years for independence software vendors. So I was much more copyright oriented and I can share with you the big difference I've lived since I joined Numergy. Open source is just incredible but what kind of workload the customer is waiting for and that's my experience. So it's about knowledge, under control knowledge and picking up what should be done to deliver the best service. So think simple, which is not the right way to think open source because you have so many options to code. So adopting for us some agility in a way that we are producing a lot of new features or we try to. That's also something we want to manage as a cloud service provider because you know that we are compared with a very large logo in US one which has billion investment per year. So we are not poor, it's about million but those million has to be revenue somewhere. So working with open source I think he was refreshing what I learned during the last 25 years. I know that everything is possible but at the end I mean it is possible for the right go to market and also if you don't think the complexity you are not able to fix tomorrow you're going to get to fix it tomorrow, after tomorrow. And as soon as you launch the service and customers rely on the SLA you should be very serious about that. So what about you in terms of the other options you looked at before settling on OpenStack? So we had before settling on OpenStack we had our own inbuilt in house technology but also when I joined we looked at Eucalyptus and CloudStack and initially just started with a paper evaluation setting up some criteria that we thought was important to us of course momentum and openness and those kinds of things were a big part of that. The architectural quality as well and the direction was very important to us and so after that paper evaluation we did do sort of hands on hey how does this actually work in the wild and we did that with CloudStack and OpenStack and at that time and this was back in the grizzly time and just before that we were just a lot more comfortable. So CloudStack won? Absolutely, absolutely. And I'm just here to convince you guys all to switch but no, OpenStack won and obviously we've been both pleasantly surprised with the continued momentum but also as Patrick mentioned it is sort of a journey to get to the point where you're ready for production workloads on top of this and the open source aspect of it but beyond that even the open community really important for us and you know from the storage you know we're using stuff on the networking side we're using Conrail and all of these things have come together for us I think quite well. Good, thank you and if anyone in the room does have questions please do wave your hand if I can see you through the glare of the spotlight I will draw you into the conversation and while we're going on one of the things that a lot of enterprise clients especially are talking about and grappling with is this question of whether they take OpenStack from trunk or one of the distributions or a distribution and support what did you two decide to do? Well for us we're using the packages from RDO but it's pretty much our own what we put together sort of shift cookbooks and we've invested a lot of time in our continuous integration continuous deployment technology so that we know that as we take up new elements whether it's from the open source community or from our own contributions we're able to test this in a way that we have a high degree of confidence that we can push it out there similarly I'll add that within our data centers leveraging availability zones quite heavily so that our application developers and our platform teams that work on top of OpenStack are able to take advantage of the availability zone paradigm so that if we screw up and take an availability zone down which is going to happen that their service doesn't go down and they don't suffer sort of an outage. So for us again I will insist because we are a service cloud provider operated only in public through partners security is always the first topic to cover with any partners or or customers you know that Numergie is born from a public initiative, it's called Andromed it was about three years ago and the idea for the French government to say to the market then we need some actors we need some data centers in France we need to protect data because some could be export but some are not able to be export for many reasons I'm not here to explain why they are okay or not but that's a concern, that's regulations so we should assure to our partners then we are very serious about security so we build the security operation centers to protect the data centers but the platform also everything has been encrypted everything should be VPN everything should be load balancers so that's a very classic concern for any type of customers but the first topic we should prove and that's why we are running this year I mean the certifications to get the 2701 but also the CSR stars that's another way to say what I said it is writing, it could be checked so security it's a big big issue when you are public platforms and especially if you display on the flag that this is open source because for a lot of security officers open source means backdoor they have the same feeling about the corporate but they have to accept it because they cannot control so at the end I mean security is it a concern? of course it's about confidence it's about confidence you bring to customers but you bring also to partners that's another way I mean to be compared regarding the others players on the market and I think the merger did a very good jobs in terms of security we continue to invest in security and the French security offices are checking our infrastructures not to check if we get backdoors but if we are very serious in terms of security vision, approach process, techniques software, equipment data centres all those layer has been scrutinized and right now we are very secure but as soon as you say we secure I mean you are tired a lot of people which are looking for is it true yes or no? it's a challenge isn't it? so keep you alive every day and sometimes you will start to sleep at night also so for both so for both of you then you're saying that you wanted to retain the control and the knowledge over the code that meant you wanted to do it yourself rather than taking a distribution or someone else had taken on those steps no we were contributing to Solum we were on the top one just before Rackspace so you know that Rackspace has built OpenStack so as a very small company born in Europe we are very proud about that but Solum was heavy attractions some months ago and is moving so I would say OpenStack or Open Source is every working ways that you too selling that it changing very very often and as a CTO when you said I'm looking for for the next three years you should keep up the ship where it should be with the right cap and that's something you need to mitigate every day with your teams with the providers and also about what you learn because on a journey we learn a lot of things so we know that because you are more and more popular I mean the attacks is coming from very often I can share with you that we are storing 4 billion message per day only for the security the good things we're going to do big data on that and we're going to sell out services about security so that's another way to learn on a journey that things you are investing could generate return on investments but you should be patient and the board of the writers should accept also that's another part of the job so that worked a probably very similar in terms of having probably not necessarily control over what's going in there but we have a lot of best practices and a lot of existing technology at work day in terms of how we manage our data centers presently and so a lot of this work is integrating OpenStack with those existing things whether it's telemetry or logging or security appliances you name it there's a lot of stuff already in place and so there's an imperative that we integrate this technology with those existing technology but also to operate this you need to have a high degree of familiarity with what's happening beneath the covers we don't allow anybody outside of the company inside of our data centers and so something breaks we've got to be able to fix it and so that traditional path of ramping up on the technology figuring out how it breaks and how to tear it apart was a critical element in the maturation process for us so for both of you that path creates an immediate challenge which is skills where did you get them from did you train up existing staff or hire new staff so this is a brand new team at work day we did leverage some of the existing in-house expertise where we're applicable but just hiring new employees and investing in the new employees making sure they're as you'd expect smart high bandwidth that can learn technology quickly are comfortable and flexible with requirements that might be vague and able to learn with these learning technologies able to contribute maybe not right out of the bat but within a short period of time so what do you hire do you hire an open stack engineer or do you hire someone who understands asset of technology our focus has been on software engineers and software engineering people with computer science degrees we have a lot of sys admins and people familiar with IT infrastructure at the company and then bringing on software engineers with that sort of sys admins and bent was the path we chose Patrick we were following two roads first because we started the company so we were looking for to hire people as I said the availability were very rare in terms of real open stack experience we found some people which has been experience of open stack ccs that's okay that's a proof of concept for yourself when you are home and you are thinking that you could do something interesting at home but it was not at the right level I mean what we were looking for so we were forced to looking for people then we gonna train and I was loving this period because we were picking up some doctors coming from when I said doctors we respect because they have a very large brain and they can read, learn and express and train I was following what I learned when I was in US I mean train a trainer that's one way it was enough to build an architect teams about seven people but you have to find a very senior one which has not a dreamer it's over it continues to dream but he's dreaming to build a team much more than getting more value on the market so that was one part of the planning we were thinking about and we were doing but the other ways we were working were we were deciding to talk with some school engineers or engineer schools especially for the last four years before they get the degrees or the certifications so we were building with the team the cloud certifications from some specific schools we have around Paris two of them and the goal to get those engineers because we were showing the project we were training them about cloud not only especially open stack and that's another way to getting the best of brain in terms of new engineers because they are still dreamer and I'm going to be a dreamer again but also I can tell you to manage those people you should to relearn all the time what you learn what you think it is true because the right way to manage those new energy it's a big gap for company so you should accept that they are coming from with skateboard you should accept that they have doing some flash ball I said not dead behind that but sometimes I mean during the day I mean they can get up and say okay we are going to play for 10 minutes or 20 minutes so the breath for those time of team is very different and you should adapt what the board is thinking about it and the right way I mean to put a new atmosphere where you get the best of the innovations when you are senior as I am that's a new lesson but I like to learn and it was a very very good pleasure I mean to work this way so we blended some seniors, some doctors some brain new engineers which are not be yet certified but many of them has been joined in numergy that's another way to get your people very loyal and that's something we need to build a culture because everybody is thinking about Silicon Valley but you can do that I mean around Paris and that's a very good experience returning experience I'd add to Patrick mentioned with interns at least that's what we call them in the US and new college grads we've had a lot of success identifying them from some of the top schools bringing them on and that has a snowball effect as you bring these students on and they go back to the university to get additional candidates coming at the next opportunities and that's been a tremendous experience for us we're a start-up in Switzerland we're going to be providing public cloud as well we went live beginning in November and we're looking to make our cloud GXP compliant and I was wondering if either of you have done that and how difficult it was to get there I'm laughing but it's serious not for us which part of Sweden are you on why are we in Basel ok so I'm currently talking with some bodies in Zurich but it was not for the same subject you just want to get our our open stack platform oh no we've already built ours it's ok we're done but we didn't get to this level probably we need one part of let's say the radar we want to be certified we want to certify for France the open stack with a classroom with a very high skills to get these certifications as a big player did for many years I mean I think about Cisco and others and that's another way to bring the people and getting the confidence from them so we are thinking seriously about that and that's something from the market perspective I mean if you want to see switch the big bang on Twitter and others they have to think that the training course the certification and the platform and the player are available because they want all in one answers and that's somewhere the strategy following for us for my English that's ok it hasn't been a focus for us no ok something you feel you have to have definitely in Switzerland it's pharmaceutical companies and they're not going to come in our cloud if it's not GXP compliant we've got all the training that's all sorted we're moving on weight we've got ISO 9000 and one sorted we're going for 27000 that's almost there GXP hopefully next year but I was just wondering if anybody's already done it because there's not a lot of people who have and so it's a little bit of a blind alley we're sort of crawling along in the dark so when you've done it come to Vancouver and tell the audience about it hopefully yeah I had a question something of course to pop to my ear when you said cookbooks and two questions for both of you but the cookbook one first you hopefully realise there are open stack cookbooks out there in the community in Stack Forge right that are now maturing on one of the course so I have to nibble on that a little bit have you ever tried to contribute back to it and who might be doing that and then just in general you've taken open stack I'm kind of curious what type of what type of parts have you done for giving back to the community I'm just kind of curious how that's scaled in both your companies I can start with that we started with the rack space cookbooks where that's what our earlier clouds were based on we did go down a different path in some of our production deployments we do have an initiative right now to use the Stack Forge cookbooks and I believe we've made some contributions back already but we're just at the early stages of that and our future production clouds are most likely based on Stack Forge I am curious to see whether or not any work is being done to do Stack Forge with triple low just because of my own belief triple low data center scale installation and upgrades somehow that should all roads should go through that someday but as I said perfect I think I answered the question so as I said we were contributing to Solom in terms of cookbook I'm quite jeopardized because when I was talking with my people a year ago everybody said puppet six months later they said chef and right now they said Ansible so sorry to share with you one secret I got a musician so that's only way to writing music so I like to change technology but at the end I think about three years if I change all the time I mean what I add in it's complexity for my colleagues are working in productions and that's my big concern so I don't care about Ansible or puppet or chef I know that we did it as soon as we have to contribute or to reverse what we develop but what I did in terms of internal rules every of people of my team can contribute but they have to contribute in the name of numergy and that's something you need to balance because people are dreaming about themselves which is cool but did they have to do things that the company is looking for to develop? yes or no so I put it some rules I learned in my copyright word I said you contribute in the name of numergy means that you have ethics wide way to document and also that's another way to say that numergy is an innovative company energized but the good engineers but this is not an individual contribution so puppet, chef I don't know and I'm waiting the next one and it's confused somewhere what is the best for the next three years that's a big question you have to ask so when you are swimming in the swimming pool of open source for years you don't see any issues but what I see if I'm thinking in the next few years I have to keep up the code which has been developed 10 years is it something you think about it when you are doing? so this raises a broader point that we wanted to touch upon we only have a few minutes left but you talked about the move from puppet to chef to ansible and also at the beginning mentioned the issue with open stack release cycles they're rapid every six months something new arrives it's okay perhaps for a developer it's okay perhaps for dev and test it raises issues for both of you in terms of running a mission critical dependable product definitely I have no words to say it's interesting for the test and dev for sure but how many company as developers in ours have their own applications because most of them have been moving in offshore it seems that the developer is now coming back on top of the concern but we have to help them to pick up the right tools for the next five years and the next five years with every six months on your release it's a very difficult forecast so you should be in that's why the energy is one of the contributors and we pay to be part of I put my engineers or the team on the key critical compartment we need to operate the platforms so security network, storage and of course CPU or calculation nova and also the storage it's a one piece we're driving you to the big data area space that's also a big concern you need to balance what kind of technology not only for the timing but if you have to move 10 or 20 peta what time I need to move from one way to another one which is the same data which have to have the recycle to put it in a new web because you want to extract the value of so that's a concern you have to have to fix in terms of CTO because your role is to define what should be done how we can do that with which budget and at the end for which customers and are starting always by customers I don't know what the customer want to do with big data but I know that they want it that's a common respond to think everyone wants big data but they're not totally sure what they want to do with it the six month recycle doesn't bother us as much our current strategy is probably six months behind the latest release once it's been vetted and had a .2.3 sort of upgrade and bug fixes and things along those lines and then we'll let it incubate through our continuous integration process and then go from there good thank you so we are I'm afraid out of time I would like to thank my reduced panel that they did very well talking for what they thought they thought they would have half the time to speak that they actually had so they filled the space nicely thank you to all of you for your attention and for your questions I hope this was useful to you and Alison will give you vouchers for t-shirts if you want the one on the way out of the door so thank you all very much for your attention