 Okay, thank you very much everyone for coming very happy to be one of the first breakout session so Question I'm going to try to answer today is our enterprises ready for the open-sac transformation lots of people Think it's a weird question And I think so too, but we'll see that it does make a little bit of sense so first of all, let's let me Start with a little introduction on who in advance is In advance years a company started in 2008 It's been growing quite fast since it's selected open-stack as its cloud computing engine. We went from 20 people a year and a half ago to 120 something Today and the gross is still going so we do accept application if you want to come work with us We've got plenty of customers big names smaller ones We make a little bit of revenue which is lucky And we are one of the top ten contributor of open-stack and we've been so over the past four releases about me I Work as VP products that is in advance. I also take care of the pre-sales team I've been working on open-stack since Day one since it started being called open-stack The Austin Summit. I founded the open-stack telemetry with a few other people Also known as Cilometer and I've been traveling the world doing open-stack and mostly talking I mean, you know VPs that they don't do real stuff anymore And if you want to send me a tweet say how bad the presentation was at Nijaba, that's that's where it happens So a little disclaimer. We didn't hurt any enterprises doing this presentation And as you'll see a lot of what I'm going to be saying here today Is not really completely specific to open-stack. It could apply to any open source cloud technology So the first question I've been asking myself is he is open-stack a product and I don't know what you guys think. Is it a product raise your hand if you think open-stack is a product At least we have one person so he's going to be able to contradict me because obviously I don't think it's a product it's From my point of view a technology framework and since it's not a product comparing other products with technology Is a little bit like if you were comparing a car and an engine. Yes my car has an engine But my car cannot be compared directly to an engine and clearly What you need to do is to keep focusing on your use cases Whether Open-stack is Enabling your use cases is really what you should care about and you may want to Use other components in order to enable all your UK use cases for example today open-stack offer a great API to Access billing information actually measurements But it doesn't do billing by itself and if your Objective in your cloud is to be able to build customer. You may want to integrate with some billing software and Once you've done that you're a little bit closer to a product than when you were when you started with open-stack another example is you may want to have a self-registration portal Does open-stack offer one? Obviously not So you may need to build one or use another one and then again you'll be a little closer To having a product So if you take open-stack all by itself, no, it won't ever be ready for the enterprise You won't ever be able to drop open-stack somewhere and Say hey, it's ready for my enterprise yet Enterprises may want to use the open-stack engine to build their clouds It's not because just the engine Doesn't do all that they want that you may not want to use that engine to power your car and Why would enterprises? Be interested in the cloud infrastructure Well in summary, it's very simple people either want to save money or to make money when they say save money generally it's simply either by reducing the operational costs of their data centers or To have some flexibility switching between CapEx and Alpex When people Want to make money? They may want to use a cloud to reduce their time to market They may want to enable new initiatives or they have big plans They want to go to a very large scale and they really need something To help them achieve that So when we're talking about saving money It's very tempting because we are in the realm of free software and that in English There is no distinction between free as in free beer and free as in free speech Lots of people will say oh, it's free software. Therefore. It is for free. It's gratis, right? So I'll be saving on licensing costs. This is why I'm going to be using open-stack little poll Do you think licensing costs are? Good enough to go to open-stack raise your hand if you think it is Okay, you know the trick now. Oh, no, there's two people. Hey So the others you say no correct raise your hand Good, I like you just one you can leave So what I did is I I talk as I talk with quite a few customers one of them was kind enough to share with me his cost breakdown of Operating a virtualization infrastructure, which is something really close to what we're doing, right? and I was really interested because He was using VMWare. So therefore he was paying the big bucks for his infrastructure and I was really surprised to realize that he's licensing costs Where only at about 10% of the overall recurring cost for his data center? This is clearly something that is showing that the number of Humans that are needed to operate a data center Have a much bigger impact on your cost than the licenses this is clearly showing that Licensing all by itself May not be the thing where that you expect in saving money So just a little these little small sentence here Discuss also of course vary a lot depending by region. Why that? because The the price at which you pay for people is not the same because the price at which you pay for power Is not the same so you you can expect very big variations on this breakdown From one region to the other don't take this cost analysis as your cost analysis This is just valid for the one customer in France That revealed these numbers to me, but still I think the differences are so big But I think it should apply to let's say the Western world in general pretty well And let's face it free stop software still has a cost I mean you may want to Have maintenance. I mean new versions of your packages being delivered to you And whether you're going to do that internally or if you're going to pay someone to do that for you this has a cost expertise this is Clearly something where you have to spend at the big bucks to get the open-stack experts There's there's not enough of us So we are expensive. We're happy about that because we make a lot of money, but still that counts in the equation and To build up these expertise internally you need to spend a lot on Having your teams learn about open-stack. This has a cost as well So here is the answer. Obviously. I don't think that the simple subtraction of the licensing costs are going to be Enough to justify switching to open-stack So this presentation is kind of weird and we are at the open-stack summit and I'm saying no You shouldn't expect open-stack to be saving helping you on saving licensing cost. Well, obviously I Have a trick up my sleeve right and the question is if I don't save on licensing. Why should an enterprise? Pick open-stack. Well, there are quite a few reasons just for open source or free software in itself The first one the one that for me counts the most is the adaptability If I'm missing something in a software and it's built by a proprietary vendor, I Can cry I can yell I can dance I can run in circle But the product manager of this product Will in the end be the only one making the decision of whether my missing feature is going to appear or not Which free software? That's not true security I Mean, let's not compare Whether my stuff is better than your stuff in terms of security as everybody knows security is not a state It's a process, but in this process even the ability to review Some of the key component that makes a security of a product something essential or would you rather? Delegate your trust to someone else. I know someone that wrote a great paper on trusting trust and I'll let you Think about that sharing So we can share a lot of things But cost is one thing that I love to share If I'm not the only one having to pay for something If I can chip in a little and have the thing move forward That's a lot better if I've done if I have to pay for the world development of the thing, right? Standards Well, I really believe that a Good standard is at least a standard that has one implementation that is done in open-source fashion if you've got to pay for some kind of patent to have access to a standard or Buy a product to get access to this standard. This is not a valid standard to me hybridation How can you pretend that you're going to be doing hybrid? if two things are not operating the same way and Given the ability of free software to be used in different ways by different people But yet provide the same API This is what is going to allow me to have flexibility in the way I deploy on the cloud, but of course the the benefits should not only come from The cloud itself But it should come from the transformation That is happening on top. Let me give you some examples If I enable my business units to be quicker to evolve In a market that is speeding up Am I not in a better position to compete if I enable my development teams To be more productive productive and more creative and more autonomous Am I not going to get better software? if I enable cloud-ready applications to be more scalable Then again, I'm going to get some benefits out of it clearly all this relates to the possibility for us To be quicker smarter in the way we go to success with our products and this is what You should be expecting from an open-stack transformation standpoint. So Do I just have to install open stack and I'm done? Well, again, open stack is just a technology It's really a tool to enable the transformation Open stack is what I would call an agility enabler and I know Alexine's in the room somewhere, but with these lights in front of me. I don't know where he is We could have a discussion the two of us for ours About what really agility is but I really think that in order to implement agility you also need some tools and Open stack can be one of them Something that you're going to Use to enable your business units to do their job better your development teams and even your marketing but That's really only happen If in the process of installing open stack You change away People have to interact with the data center if you leave the things as is Trust me You'll get zero benefits Which ever the brand of your cloud? So I gave three example. Let's take a look at What they do what those groups of people do today and Where you need to bring them say Let's say I have a business unit that Deals at selling yogurt today when they launch a new yogurt They need to tell it a I'm going to need this new website talking about my yogurt or I'm launching this Big contest I need this application to be running and then they've got to wait for approval wait for it to implement their need of Course complain quite a bit whenever something fails That never happens, right? You're going to rely on your internal IT for your application SLA so basically If you need something better different or you don't care about Availability, this is not directly your responsibility and The application regardless of what it is is going to be Applying the same global policies that have been defined generally by the enterprise tomorrow I Would hope that if you want to have a successful cloud implementation these guys would look at the problem a little differently First of all, I really really believe in the Benefits of iterating The whole Idea of life if we look at it from a genetics perspective Is a succession of trial and error? Do you think that the system the big system that we are? the human that we are becoming out of this biology is Any different From the biology that makes it I? Think we need to be able to try out things and figure out where they fail and fix it and iterate and ask the quicker You iterate The faster you reach your goal I have a friend that works on a very small creature called colambol colambol is About one millimeter on average in size It's been existing for a very long time I mean we it's a little creature that has survived all kinds of cataclysm and The great thing is that you can find colambol anywhere on earth These things have a Lifespan of less than a day They are the most adaptive creature on earth. Maybe not the most. I mean I'm not an expert, but one of the most and Why that it's because they are in very vast amount Reinventing themselves at each generation trying out crazy things Let's have wings. Let's have a fifth leg. Let's have two brains and Whatever works survives and it becomes better and better and the funny thing is that if you look at colambol In Amazonia and the one in Africa They share lots of genes But some of them have wings some of them have tails and They are very very different creatures When you look at what they are and they adapt to their environment very well And I really think that in our data centers, this is what we should be trying to do We should be enabling people to fail fix and Test again and maybe fail until this is seen in order to do that The people that are going to benefit from an evolution Should be the people that are controlling What they need it should be the owner of The application that should directly Decide how and what is going to be tested next? You can't do that remotely you want to be able to define a minimum viable product and test it as quickly as possible and In general One is always best served by itself, right? So if I want to have a very very highly available application Who is best to operate this application but me? Of course, I can't do everything But let's lower one layer of Who is going to be responsible for my slA? I'm going to be responsible for the slA of my application. The data center is going to be responsible for the slA on what's underneath and Really by providing them autonomy, we should be able to allow We should we should allow them to modify the the policies so that it match their needs and not Dumbly apply something to them that doesn't match a need when we take a look at the same thing from developers perspective Right now most developer received lengthy specification that they try to understand and They will develop something that may match What was expressed but generally, you know, there is this little problem that is called communication That makes it that two people hearing the same sentence won't understand exactly the same thing Then the developer is going to have to fight for some hardware location Because of course, you know the resources are scarce. You can only fit a few servers under your desk and as a developer generally will your first incentive is to Deliver the thing to the production people so that you don't have to worry about it anymore So clearly you're not responsible for the slA of your application. It's a people the operational people that are going to be maintaining the application into existence which are going to be having this burden and in general I've been told that All the developers should be using this data store because this is the official data store of the enterprise When we switch to cloud we want people to become more Developers and operators what we call DevOps, right? And we want them to be closely interact interacting with the stakeholders. We already explained that second ago, of course Deterration principle should happen as well There is nothing better than being able to be responsible for the hardware location that you need You want to have the same team That is developing the application to be operating the application because this way when there is a problem it can be fixed and It's much quicker to understand where the problem is coming from when you know the application how you develop the application and You can build the slA into the application and You can go even a step further once you're done with that you can build hybridation within the application And I believe that's the only way you can do real application Ibridation if you want to have the same application running on multiple data centers if you haven't sold Where and how your data is going to be replicated across at the application level Trust me. There is no magical solution for your data to Move across the center at the moment you want to burst marketing so basically another group of people that tell it what to do and In general, they don't think it is quick enough so they're going to be subcontracting to some agencies to develop application like crazy They're going to be complaining whenever the deployment of the application fails whenever something fails They don't understand why it takes time to do things because they've told you what to do and They seem to always buy their Advertising space before you've told them how long it would take to complete the project Of course at some point they say oh, but the heck I'll pop out my credit card and I'll go on AWS So that's very simple to track what these guys have been doing, right? In a transform marketing if they have the ability to allocate their own hardware on the enterprise last enter if they can provide Access to these provision hardware to their agencies if they are in charge of their operations if they have to fight with their own delays if They are able to Get their application in front of customer Through a wall chain of responsibility there's they'll need to find someone else to blame They also will become traceable You'll be able to identify What has been spent to do what? Because everything that is happening in your open stack cloud is measurable and You can attach spend to a tenant ID or to user ID and create all kinds of reports That allows you to go into very Much details about your IT spends So what you want to do is globally transform The interactions within your enterprise transform the complainers into creators Transform the consumers into the operators Transforms the development diva the guy that say oh, I only code I never operate into DevOps And clearly try to transform a large main month Into a lean organization that's operate efficiently But of course you have to operate that transformation for it to happen Clearly computing is all about raising a fraction level. I Was born after the beginning of computing, but in the past 25 years I've seen the same pattern happen again and again we go from Binary to assembly to language a to language b and each time we can do the same things quicker easier and we use these abstraction level to find our own gravitational laws basically when we scale we have to fight against Data gravity we've got to fight against Application states we've got to fight against lots of things so we've been Changing the model trying to adapt it For it to make more sense to solve these problems. So we went from mainframes to PCs Then oh, maybe we still need central data. Let's go for client server. Oh, this is this is good But this is not enough. Let's go to a three-tier model. Oh But it would be even better if I could do distributed and once I've done distributed. Let's go to hybrid If you look at the processor level that the evolution is about the same we went from a single-tasking Operating system with processors to multitasking once then we switch to multi-threading multi-core and now Using cloud computing using OpenStack. We can do multi-node You don't that amazing so clearly cloud infrastructure is just one more abstraction layer one that allows us to Instead of caring about individual computers care about millions of nodes. I think there were a single entities This is what has been described many times as a pet versus cattle discussion But of course in order to benefit from this you also need to adapt your tooling You need to allow to change your centralized databases into distributed one to go from virtual vertical scaling models to horizontal one file systems are Never the best choice from my point of view. You may want to use objects or blocks Go from single allocation of work to bus System to allocate work et cetera et cetera all these are transformation that you need to enable into your applications So if you go and say let's use the cloud But let's keep on using that all big Oracle database that I spent so much money to put together This may work, but will not necessarily provide the best results in terms of scaling and really OpenStack has been built For this transformation to happen. It doesn't provide all of them, but the real purpose This project was started was to enable just this Another thing that I see people trying to do is to migrate all of their application at once to OpenStack but the problem is that OpenStack is not meant to host this application without lots of transformation and Even if you don't do any changes to your application the idea to move it to the cloud Just so that you have a unified data center with everything in it and you feel very powerful has a cost and No real additional value provided to you what you really need to do is to accelerate What's happening not just create more work for the sake of it, right? and one thing I've learned That's coming from the Unix era do one thing and one thing well Lean enablement cannot be done with bloated software if we go for a race of Having all the features in OpenStack will have bloated software that nobody can be really satisfied with So how do I go about doing this information? The answer is for your specific case. I don't know but one way I sometimes suggest is To start small start your cloud with new applications or components converts your existing application as They happen to go for a new major revision in their normal life cycle You can mix and match Technologies you can still use this old Database that you were using before and change the front-end for a while that work That's a good way forward You don't need to have everything to be cloud at once and That's using this that you should start seeing the real benefit of cloud Converting your existing application won't show that Once you've done the first step, then you should try to work on making hybrid application Starting by the front-end that's the easiest part making it work from multiple sites at once and then Think about how you can enable your data to be replicated on multiple sites all the time And that's when we you will see the benefits of OpenStack hybridation because having the same Environment run on multiple locations maybe not maintained by your enterprise but Delivering the same value this is Really where we are all to be all aiming at the moment And then you've got to change the way your IT team is delivering It's really important that each group within your IT team Be responsible for delivering the resources globally If I am not seeing the end result of the work I'm doing It's going to be very hard for me to feel responsible of anything, right? and Let's stop dividing the people that design thing from the people that operate things because There's always great ideas, but as long as you have not implemented them you have no idea how they perform You want to stop doing the silo thing? You want to define? Agreements service level agreements based on API and measure Your API responsiveness or your API availability You also want to stop completely the pre-controlled craziness It's not because one or two individual in your organization are going to be Weird doing stuff not the way you would have expected that you should force everyone to go through a three-step human process to launch one VM And believe me I've seen that happen trust is the very fundamental thing about delegation and and once you have proper delegation then You have what it is and what is needed in order to dispensabilize people to provide them with autonomy and In turn they will be created You shouldn't care about preventing from things happening control measure things and control afterward I mean You know if somebody has a crazy idea That is bad for your company It's a lot less often than good ideas that are good for your company So let's enable the good things and worry about the bad things afterward really Measuring your end goals Measuring what you are providing in term of value is the thing you need to be concentrating on This is really the only thing that matters So to summarize OpenStack is not a product It will never be ready as For the enterprise by itself License cost is not what you're going to be expecting out of OpenStack or should not be you should concentrate on enabling better processes and better tools and You should be transforming your enterprise to produce more value Thank you Done now. There is a microphone in the middle here if you have questions Feel free to ask them right now Can you talk a little more about your vision for how databases should be because I represent companies that have Mainframes and Oracle databases So The problem is that if there was only one answer there would be only one database Unfortunately, depending on the type of data and the type of transactions you're going to be doing You have to have to adapt Your choice of database to it and not the reverse I'm not going to be dealing with time series data the same way I'm going to deal with reservation booking for an airplane so The only thing I can say is They are now more than 200 different databases They are mostly free software. There is I think the right one for you need and Data architects are there to help you to to pick the right ones Any other questions? Well, if you're too shy to ask them in public. Oh, go ahead This really depends of what your end goal is if you are going to be Transforming your business into a service provider That is going to make money out of open sack I would recommend to do it yourself or something close to it You want to get the knowledge of the key tool with which you're going to be making money, right? This is your real production environment and you don't Delegate the knowledge of your production environment to someone else. It's so key to what you're doing but if you are an An enterprise who is doing yogurt open stack is just a tool and Generally, if you if you are not going to scale to more than the size and servers The build versus buy is going to be in the favor of buying It depends on who you buy from and what they're delivering to you if they're delivering to you something that is 100% open source if they make modification or add Component to it to do it the proper way without forking open stack You have no real luck in I mean, of course if you take something that Has some proprietary software in it Forks open stack from what it is then switching to another vendor is not going to be possible but think about your reversibility if you Have the guarantees the right guarantee that everything that has been delivered to you is completely reversible is fully accessible to you Trust me you won't be locked in but you have to be very vigilant in doing so. Yes Yes, that's very key the understanding how to maintain their logical environment from a to z Is what needs to happen? Yeah, so there is one very misleading concept in the use of the word DevOps it sounds like it's one person that does everything We're talking about teams here You want to assemble in a team the people that have the Some of the knowledge needed to develop and run the application Design build and run should be done within one team But of course you still need to have specialists a specialist of high availability Together we as a specialist of deployment together with a specialist of coding whatever you're coding Okay, it's not just oh, let's put that on these poor guys shoulder and he'll do everything by himself This is the the the one man operation thing that we are trying to avoid at all costs, right? It's actually I Didn't represent it, but you're going to be fusing from three groups Who owns the application? We's asking for it the stakeholders of the application Who operates it and who? Build it and this should be considered as one team And until you've got all these three groups in one team You won't get the real results as benefit of the cloud Okay, I think I'm out of time. I'll remain available while I'm exiting the room. Thank you very much