 Thank you everybody for coming. This is the first session of the Evaluating of the open stack. We have a session every day Two or three two three sessions of the track. So today we'll start with Future of the open stack distribution Luckily, there will be Two sessions today. There will be three sessions tomorrow the same place at two o'clock and then every other day will be Morning sessions on the Wednesday and Thursday. They're all under one track evaluating open stack so all of them are related to Help you to make decision of On the open stack distribution what features of open stack and what are the things to look in the open stack? So without further ado, the first talk will be on the future of open stack distribution by Mark Smith and Thanks Arkady Thank you very much Arkady The pleasure to be able to talk to you about the future of open stack distributions. My name is Mark Smith I'm product marketing manager for Suzer open stack cloud Pete is the director of product management, which is why he's wearing a shirt and I'm wearing the casual clothing. I I'd like to start just talking a little bit about our The importance of choosing and making decisions right so quote from A movie you must choose but you must choose wisely So we'll face with choices and one of the the things that we appreciate about open stack is that the choice of consumption method is within our control but also like this quote from Peter Drucker who's a business analyst and an author He said this whenever you see a successful business someone once made a courageous decision Now quite like that we're in boston Historically, I'm talking to you as a person from britain, right in boston the right to make courageous decisions is baked into the history and The right to make courageous decisions is baked into the history So if you're an open stack user, you've already made your first courageous decision Or maybe you're just looking at it and evaluating things right now, but you've chosen open stack You have more choices that are on the way So let's just start by talking about consumption models And now the user survey this year from april 2017 Has a couple of comments from users That re-emphasize the fact that they appreciate choices within their control They can choose how they consume open stack That's that's really important. Now you Before we we talk about the individual choices We're here talking to you From the open stack perspective As soon as we don't have an axe to grind on which option you choose Now the reason for that is that we power a lot of managed cloud offerings We obviously have an open stack distribution and that's the theme for this particular discussion But we also want users to be doing it yourself building your own Now that's really important because Education getting familiar with using open stack is absolutely critical And a lot of customers that start by testing from a do-it-yourself perspective End up moving to a different consumption model as they Mature in their requirements and as they make further choices So what are the choices? Well a do-it-yourself approach a build your own Means that you're going to take the upstream code You're going to decide how that works in your environment And you're going to take on the responsibility for building your cloud managing your cloud upgrading your cloud And we're going to a few more bits of detail about that now of course if in a do-it-yourself Build your own environment you could Take the open stack distribution code. You just have to support it yourself But then you'd get an install framework that kind of goes with that With an open stack distribution what you're getting is a pre-packaged Version of open stack with choices that have already been made for you With a deployment framework that should make it a lot easier to get started and get up and running So again, we'll talk a little bit more about an open stack distribution A managed cloud is really where You're taking the least amount of personal input to the cloud. You're basically saying to someone else. Hey Drive my cloud for me I'll pay the taxi driver. You take me from a to b You're responsible for administering the cloud and then actually doing the work with it But you're getting somebody else to build it for you manage it for you Upgrade it for you all of those things So let's go into those in a little more detail. Let's just talk about the pros and cons of a of a do-it-yourself approach now My illustration here is of a kit car, right? You can buy the components for a car and you can build it yourself If you're an enthusiast You may well do that. Can I just check as anybody built their own car in the audience? No one prepared to make the courageous decision of building their own car Not unreasonable. Okay, that's a lot of hard work. Maybe you don't like spending your weekends with your Up to your elbows in oil and And piece parts for mechanics some of us do And of course in the in the open stack world many of us decide that we're going to do that now Just to put that into context a colleague of mine Many distributions ago calculated that there were 1400 components That you had to configure To build a do-it-yourself open stack cloud platform It's probably gone up a lot since then The user survey for this year from april 2017 Shows that of the people that responded to that survey three quarters of them with Open stack clouds are using between six and eleven of the open stack projects between six and eleven Now that's a lot of componentry to build deploy Maintain and look after but here's the pros of doing that open stack is modular by nature. So You know you have choice. That's a really Appreciated and valued by open stack users. You can install what you need You can save a lot of cost How can you save cost? Well, if you're building a car and all the components are free That's a fairly cheap car What it doesn't take into account, of course is your labor and your time the value of that time to build it configure it and maintain it And everything else that you need to do but you can save a lot of cost If you have the right skill set you have the right in in-house team of people You also get almost unlimited flexibility and customization. So you can make open stack Exactly what your business needs it to be now. Those are all really positive Pros of a build a build it yourself approach Just flipping the coin over very slightly Back in 2015 451 research As an analyst commentary said that open stack is consistently recognized as being overly complex to configure deploy and upgrade That matches the experience of a lot of customers that that I talked to they they appreciate open stack and all the value it adds But many of them have had a pretty complicated time of getting it up and running And I think that even came out in the keynote material this morning Has that changed? Well, yes, it has open stack has matured. It's become easier But again a couple of commentary A couple of comments from customers in the user survey Sometimes the complexity of the projects and deployment can be a major problem So that was just from April The biggest challenge is to upgrade the production system Now if you have built your own open stack cloud that is a challenge because every six months the code base is going to change You don't have to change Every six months, but the longer you leave it the harder it will become To get up to the the latest revision So again, that's a that's a difficulty That we we have to weigh in the balance of the the courageous decision Now, of course if you do it yourself, you're going to have to deal with the maintenance and the upgrades even Of the code you're using so There's an overhead there that you've got a factor into account. So that's the do it yourself approach Let's change tack a little bit and talk about a distribution Now when I'm Buying a new car and here I probably need to say a new automobile When I'm doing that I normally go to a dealership and I've done my homework. I've done my evaluation. I've chosen my model I've chosen my color And I drive the thing out of the showroom when I paid for it and I start doing something with it straight away Now I have to say that open stack is not like buying a new car, right new car Simply drive it away and start but you can make the whole process of of deploying Getting up and running and starting to do some work with your open stack Much quicker and easier and slicker With an open stack distribution because the deployment frameworks here Suze uses crowbar crowbar We found to be extremely useful in in making a configurable open stack cloud Pretty easy to get up and running Now what else do you get with a with a distribution? Well another of the comments in the user survey was that Installing the high availability elements and configuring those for open stack was pretty manual pretty difficult to do So an open stack distribution should do that for you And it should be part of the the framework for rolling out the whole of the cloud If I'm buying a brand new car, I want somebody to have done some quality testing To have made sure that it was reliable and also to give me some warranty and some support Well, that's what you should expect with a distribution You're not doing it on your own. You're doing it with a partner and they should provide you with A quality tested ready to run Pre-packaged mix of features That will allow you to do what you need to do now, of course doing it yourself you can Have that extra customization and adaptability You should expect a distribution though to give you Almost everything if not everything that you require And then of course you get the managed automated updates bug fixes and all of those things And you should also Be expecting to get non disruptive upgrade capabilities. So when you go from one revision to another You're looking for zero downtime You're taking away the the management overhead. That's what you should expect From a distribution so effectively Although you won't necessarily just drive it out of the showroom you should expect to be able to deploy and configure in hours and not days or months so There are pros and cons here. This is going to be a little more restrictive. It's going to be A little more prepackaged, but it should be making life a lot easier Now we're going to talk about the future of Where the distribution is going for open stack in a few moments Before we do though, I wanted just to talk a little bit about this courageous decision for Whether you go do it yourself manage service provider or managed Offering or whether you go for a distribution you're going to have to think about these factors That may be unique to your organization So first up You have the same decision to make about Cost capex opex balance As actually you would whether you were going for let's say a private cloud or a public cloud, right? You need to look at whether you want to pay for Your open stack Private cloud as a service so entirely from an opex perspective that might lead you to decide That you want to go for a managed offering managed cloud offering On the other hand, you might want to weigh up. What are my capital expenditure? What's that going to be if I decide to run it on my own infrastructure? Now if you have your own infrastructure, you already have your own compute Resources your own networking infrastructure You may well want to adapt that you might well want to take that to its next level of efficiency You might have gone from physical to virtual and now need to go to an internal private cloud You've already bought the resource So you've spent the capex. You don't want to get rid of that. You want to reuse it So the cost equation here As to whether you choose one consumption model over another Will vary according to your business You'll need to think about What kind of an IT team do you have in your business? What skill set do you have? How much do you have? If you have If you have a well educated Ready to go IT team Then the do-it-yourself approach might well be the ideal one for you If that's somewhat limited Then a distribution might be the ideal choice to fill that gap for you So you can get on and do what you need to do with the cloud Without worrying about the underlying infrastructure Your management overhead might be something to think about as well So in this case If you go for a managed cloud approach You're going to need to have a vendor management relationship You know, you're asking them to drive it for you. You need to control the taxi driver. You need to have The right management overhead to deal with that situation while you get on And administer the cloud to do what you need it to do You need to weigh up your support and maintenance overhead If you do that for yourself, how will it be done? How will it be automated? What are your requirements going to need now just in terms of again context here we have we have Private cloud customers using open stack as a distribution Running fairly large clouds with just two or three people Because they can offload that support and maintenance overhead What are your project time scales? Now every business will work out how long they're going to take to get from a to b from no cloud to a private cloud And if that has to be done pretty quickly a do-it-yourself approach might be something you don't want to consider Maybe you're going to go for a distribution or a managed cloud offering because you've got to get there really quickly Again, that needs to be weighed up The last item about customization If you're an organization that has specific requirements Then a do-it-yourself approach might be the ideal thing to use And of course your influence in the upstream community might also be More important to you and therefore the do-it-yourself approach might work well there So these are factors that you would need to think about consider balance in making your decision now we're going to take a little bit of time now just to talk specifically about Open stack distributions And that particular choice element. So let me hand over please to Pete Chavik. He'll talk us through first of all the history of the open stack distribution and then through the rest of that criteria Thanks mark So I actually Have my first open stack summit was here in boston when it was For it was the Essex design summit. I don't know if anybody else was here at that point I'm making a joke earlier talking to somebody the session was at a hotel over by the by the seaport And I think there were five meeting rooms and the open stack party after the summit was in the bar downstairs at the hotel Not a lot of people involved. So it's pretty exciting times We launched our first Susan launched the first enterprise distribution based upon Essex and fairly quickly after that there were a number of other providers Um red hat There were some piston cloud, you know nebula that launched distributions about then And since then you've seen, you know, a lot of releases There have been some enhancements that Adopting a distribution model as opposed to do-it-yourself model as mark explained have made things easier. So for example Automating deployment of of a high availability a highly available control plane. It's actually fairly complex task to configure and set up On a diy basis Because the distributions make certain assumptions about how you're going to configure The environment what packages you're going to use it. It makes that a little bit easier You know following on from that for example fairly straightforward Step to add high availability for compute nodes You're now starting to see non disruptive upgrades Again because the distribution makes a certain set of choices on how things are going to be set up how things are going to be configured It simplifies the process to do the to do the upgrade Now one thing i'll do is i'll put in Put in a plug for uh, the product working group Arkady is is on the product working group as a my And one of the things we did, uh, I think a year ago Was actually write a pamphlet specifically going into more details about the pluses and minuses of distributions versus do-it-yourself versus Managed private clouds. So I'd urge you to look through that it goes into much more detail on what some of the trade-offs are But the point is You know, this is this is the plug. I'm you know, we're in the business of doing distributions You're you're not on your own. You've got a partner that is helping you to be successful in in delivering What you really need which is a functional cloud um, and it's typically You know consists of packages that the vendor has considered stable um, so they do some additional level of of testing and analysis to make sure that some of the projects which maybe Aren't quite ready for upstream or flagged as such As tech previews or whatever And then you've got back-end support and services. So if there's a security bug It gets it gets pulled into the distribution Pretty much automatically and pushed out to you as opposed to you having to track the cvs and figure out what what's next And then one of the important things is We think it can pay for itself and this is not just us. This is 451 group did some research a while ago and and I'd also recommend that you that you look up this report That is available. I'm pretty sure it's available on the open stack website It talks about the trade-offs and more quantified Levels that it makes some assumptions about the size of the cloud and what you're doing And they clearly show That if you use a distribution you can have less staff So depending upon the size of the cloud you need fewer people mark mentioned We know we know customers running open stack with Two or three people. Um, I know I know customers that are running open stack with 10 times that number or 20 times that number Because they're doing much more much more customization and much more configuration So again 451 says they believe that ultimately An open stack distribution should drive lower total cost of ownership Not only in quicker time to get something up and running but on an ongoing basis monitoring and managing the environment Now one of the things that I think you know, I mentioned A couple people who had distributions back in the sx time frame. They're no longer here And I think there's been some some buzz about there are fewer distributions than there than there were I think that's that's kind of a natural consolidation as markets start to mature But what it's what we've really seen is that There are still people that are using distributions. They're just partnering up Working more closely with one another to make sure that you're delivering a more consistent solution That meets meets customer requirements We certainly think that the distribution business continues to be continues to be a healthy one and one we continue to uh to invest in So now this is the the higher level plug for open stack and and and at some level it's Another reason to think about looking at this from a from a package point of view and that is we see that that Open stack is becoming almost an integration platform in and of itself So we have customers that are looking at it traditionally for running virtual machines But now they're trying to say have got bare metal workloads hpc workloads or specific database workloads that require bare metal servers I can I can pull those in and manage those in open stack We've had a number of customers that have already Specifically deployed open stack just to take advantage of magnum to deliver kubernetes as a service because they want to run Containerized workloads and feel this is the quickest way for them to get up and running But then underneath of it open stack provides all this flexibility for the software to find infrastructure So you have a variety of hypervisors you can use All of the software to find networking players or providing plugins to to work in the open stack Which again lots of flexibility, but much more complexity So again one of the things I think distribution providers try to do is provide a common framework So that you can make sure that all these things work together So that you can drive the kinds of workloads that you need when you need them And so we think this is really where open stack goes ultimately It is it is the future for the software to find infrastructure. It provides the layer of apis. It provides the driver framework necessary to integrate all of these technologies Not dissimilar to what linux has historically done and linux provides a set of apis upon which you can Write applications derive workloads And then abstract away the hardware layer underneath of it to some extent open stack abstracts The software to find infrastructure to make it easier to take advantage of that for whatever workloads that you want to run So that's kind of the quick overview of sort of where we see distributions going Give you a little bit of flavor of what we think the trade-offs are between using a distribution versus doing it yourself Versus using a managed a managed hoster Um as I urge you again to follow up. There's some documentation available that goes into this in much more detail But let me stop here and any questions Comments challenges Oh, we have a question. Yeah, if you could do the microphone, that'd be great What we found is um a choice of a linux distro Or it seems to be very tightly aligned with clients open stack distro So but there are instances in cases where clients want, you know, a different linux distro with a different open stack distro So can you talk a little bit about some of the challenges and opportunities there if there's work being done in that area? Sure, so I think everybody heard that but it was a question of is there, you know What's the trade off or what's the tie in between a linux distribution and the open stack that runs on top of it Partly as and mark talked about this earlier that that you lot of there's a lot of choices you have to make So for example open stack You know we first started working on a consistent of five projects But I still needed a message queue. I still needed a hypervisor. I still needed a database to drive all that And so to some extent you you have to pick a lot of things besides open stack to make everything work um and As a as a linux distribution provider, we had a set of packages to do those things So for example, we use Rabbit mq. We could view something else, but Rabbit mq was the thing that made that made the most sense for us. So to some extent, you know You don't get to choose And and so when you when you Buy an open stack distribution from somebody who's packaging up to work with their version of linux They're going to make some choices for you and it makes it a little bit less portable um To some extent the customers that we have almost it almost doesn't Matter so much because to some extent they're looking at it. Well, okay. Once I deploy This it's a black box. I'm running running workloads over here. It's virtualized and Enterprise most enterprise distributions will support any kind of virtual machine that you want running on top of those And so that's where we see the that's where we see the the heterogeneous support come in Um, but it's also a case of if we go back to one of the things you're paying for support The way any support or any any company making support Offering support to customers makes it easier as they limit the choices So there are fewer things to test. There's fewer things to break. It's just easier to You know if you just said if you if you went back to the kit car The mark showed You know, how do you how do you offer a warranty on somebody who's built their own kit car? So there that's the trade-off Does that help? Thanks Okay, if there's no other questions we can we can hang out here for a couple minutes to follow up. Otherwise I'd urge you to look at the rest of the Sessions in this in this track to talk about how do you make how do you make a decision on how to move forward with open Stack and thank you