 joining us today we're going to do interactive small group we'll try to not keep everybody from the end of day too long what we're going to be presenting this session myself John Zeno some with canonical I manage the OEM OEM and alliance and strategic business within canonical Mark Baker who's in a pence of pause is our product manager for everything we do with OpenStack we're going to take some time and talk about making OpenStack easy and simplifying the deployment of SDNs you know I'll speak for myself for those of you that don't know me I I also am a member of the OpenStack board I personally have been consistently over the position that while Neutron has some deficiencies it is an excellent API gateway and I think the SDN community has done a nice job of plugging into it and bringing all the features you need to manage the network so what we're going to talk about is how we make the deployment of OpenStack and associated SDN easier and ultimately make it easier to run OpenStack and run a network we'll be at the end of the session joined by PlumGrid pair here please stand up there's not too many people so we're also going to then go specifically into the PlumGrid SDN and what we're doing together with PlumGrid let's see how do I go next do you know that was the one bit we didn't rehearse there we go so it's on the wheel on the side thank you so one of the things we think is going on is that as we all know software as Mark Anderson talked about in a Wall Street Journal article is really eating the world and what we mean by that is companies like Uber are fundamentally changing how the tax taxi industry is working without owning any taxis Airbnb is changing how the hotel industry is run without owning any hotels how are they doing that in disrupting these industries they're doing it through software making life easier behind that is a scale out operation that's all based on software automation at the same time the technology behind is getting more and more complex is getting more and more complex because you have this world of a hybrid cloud you have on-premise clouds like OpenStack connecting with public clouds like Amazon Google and Azure and this becomes multiple regions multiple data sets complex net networks and ultimately decisions made about data where the data may be reside on-premise the compute may be on a public cloud scale in the spiky nature of somebody's business may be bouncing between you know the public cloud and back on on-premise and this is a very complex world the scale that Amazon Google and Microsoft are able to operate is really based on their ability to grossly simplify and automate and that's what allows them to do basically amazing things in terms of running big data centers with very few people so one of the things we truly believe is that to be successful in the space you cannot solve these problems by people no more feet no more hands you've got to figure out on how you're going to simplify the architecture you're deploying in this complex hybrid cloud where software is so critically important to the business you're running and figure out how to automate it and it's the automation that ultimately is the most critical part because if you're unable to automate it you won't be able to scale if you won't be able to scale you won't be able to be disruptive and you know my simple premise here is if you're unable to do these things ultimately companies will perish just like the taxi industry is fundamentally going a change that will take many many companies that have run for hundreds of years or not quite hundreds but tens of years and knock them out of business replaced by Uber which is a very elegant model so what we're going to do is demonstrate a couple of things that that mark's going to do one we're going to demonstrate how we actually simplify the deployment of open stack with our tool called autopilot oh what did I do that's me thank you that's me to go ahead thank you thank you John so as a let me just get rid of this yeah exactly there's things we should shut down before we demo so thank you as John my name is Mark Baker I'm part of the product team at canonical working on open stack and for those of you some of you may have seen Mark Shuller was session this morning and he was talking about the Ubuntu open stack autopilot or canonicals open stack autopilot the autopilot is really a big step forward in this simplification right who's a anyone here a cloud architect no one a cloud architect that absolutely proves the point right open stack cloud architects even at the open stack summit of very rare people so and even if you can find one being able to hire them or able to afford to hire those those open stack cloud architects is very difficult for most organizations and so a big step in this simplification and automation is the Ubuntu open stock open stack autopilot we've we've worked with this this tool has taken us two years two and a half years to build and that is not because we're slow at programming right we have some good Python developers but it's because a lot of the knowledge that that feeds into this tool has come from real-life deployments that we have done onsite in conjunction with customers in conjunction with partners like Plumgrid where we encapsulate we learn from making mistakes and we learn what optimal architectures and best practice looks like and we've built a lot of that knowledge that know how into this tool so I'll just go ahead and kick this off this is really a simple three stage process a setup configure install and then manage in order to be able to get going on our cloud though we need to start with that configure piece so I'm going to click on configure I've already done in preparation a couple of things here to get going just obviously installed the autopilot itself and connected that with physical hardware this is running on real-life physical hardware is actually running on a data center in Boston I say that because there may be a little lag sometimes you might clicking and things happening but I'm going to go ahead and create a cloud this is going to be a real kilo based open-stack cloud I'm going to call it let's call it Tokyo right so we'll call it region one cloud in Tokyo I'm going to go ahead and choose some options now this is a fairly limited number of options today will see KVM as a hypervisor right most majority of opensack clouds using KVM as a hypervisor today very soon though in this autopilot you will see other options probably the first will be Lexi our container hypervisor for running LXC containers we're going to use open V-switch for our networking relatively simple and straightforward today but we'll come on and show you how we can add some other options in go ahead and do that put that on a network and we'll choose some storage so any preferences storage wise just go everyone loves that and we'll go ahead and choose safer block storage to yeah why not so we'll go ahead and do that and then we need to add some some hardware to it right so we have this the hardware that's that's being exposed or surfaced surfaced into the autopilot here is actually coming from a tool that sits underneath this called mass all of this complexity is hidden from you I'm just just sort of giving you the insight so mass is a metal as a service a bare metal provisioning tool and that is surfaced into the autopilot so you know what hardware that you have available I'm just going to choose all of these in my in my region so there we go and as we go through the rest of the talk we're going to point to how we deploy other SDNs with juju which is the tool behind here absolutely mark will do but this spot right here where open V-switch is as you see this roll this tool out over the next year with different iterations what will be added will be different SDNs that will become choices so like with plum grid after they go through our review and a certification process that will become an option if we decide to do so together as will other partners that we're working with in the SDN space and those become other choices that become much easier to deploy and in in terms of deploying them as part of an open-stack cloud with this kind of click-and-go approach absolutely and that's a great point so whilst you're doing that I noticed that I hadn't got a highly available cloud nobody likes a highly a cloud that is not highly available so I'm going to add some more hardware to that you'll now see that it's highly available I have enough machines in order to be able to spread our core open stack services across enough physical machines and then to cross two zones to create a highly highly available service I'll go ahead and click install there it'll take us into the nice gray bar land it'll tell us oh yeah let me go and take one of those off yeah exactly so boom boom there did it right there is one now which one no one there we go see all the greatest rehearsing saves the selection and off we go there we go thank you so thank you Scott there's a man behind the curtain who's not actually behind the curtain helping out there so as John said the the magic that's underneath this the magic behind the magic the magic behind this is something called juju so juju is a modeling tool that allows us to be able to model complex application architectures and do so in a simple way but choose then to deploy them to multiple different as we say substrates now in this instance is deploying it to a substrate that this is bare metal so it's deploying physical services onto bare metal and it's deploying those open stack services using the choices that we have made in order to be able to but we can also model other complex applications and deploy those too to give you an idea about juju and how we can incorporate different sdns into this environment this is an example of open stack deployed in conjunction with open v-switch so which we've just done is this the ODL model ODL over yes thank you very much so this is done in conjunction with open daylight the services that you're looking at we take a very simple example we're going to look at rabbit mq you'll see that this we have one running instance of rabbit mq we're going to take a look at the charm the thing that sits behind it this is the configuration of that the config settings that we have and if we wanted to go and see the particular implementation of that so these are services that as we refer to in the juju world as being charmed and the charm services this logical view that we're looking at are connected using something called relations yeah and one thing to understand is that each charm is fundamentally the encapsulation of the essence of an application so when you look at the sdn pieces here in essence what in the case of ODL we did in the case of plum grid plum grid did we captured the essence of the application and then what juju worries about is the connections to the other application so as you can see and you will see that the sdn charms map into neutron right and the other connections that they need to kind of correctly map into open stack absolutely john thank you so the great thing though is that we have a set of what we call a base bundle charms we can be our connecting we can save those as a bundle we have a base bundled for open stack but then we also have bundles of open stack in conjunction with plum grid for example as an sdn and other old sdn's too so if you as an end user want to be able to choose deploy open stack you can do that in a base configuration and then either add in an sdn plum grid and as a great example here and this is a an example of exactly that an open stack bundle deployed in conjunction with plum grid plum grid is a partner that we've worked with to charm their application their sdn technologies and to be able to easily connect it and this open stack cloud and this bundle represents the work that plum grid did in configurating the open stack bundle with the sdn charm that is plum grid and why don't we just do a quick check on how are we doing with deploying the cloud that we've been working on it's it's running pretty slow there at the moment five percent so a few a few minutes is a long time on your own stage but it's not when you're deploying a cloud what we're going to do is check in periodically because when in the netherlands I did this this demonstration it took about forty minutes to an hour and we won't have enough time and nobody wants to stay here that long at the end of the day but what we'll do is do a couple of checks to show you that we're actually progressing in deploying an open stack cloud that's reasonably complex with a fair number of nodes in the course of thirty to sixty minutes something that without the benefit of mapping the bundle already doing the work in terms of this example with plum grid you know in terms of mapping plum grid to open stack and the elements of open stacks being mapped together you know would be quite a bit longer I think we're at a good point pair if you want to come join us and we can talk a little bit more about well let's just stay on the both the bundle and the work that plum grid did but pair maybe as a starting point you can explain a little bit of what you liked about you do the problem that you're solving that you've encountered in the the marketplace and maybe also talk a little bit about how you think the plum grid SDN you know what's its strengths and what customers like about it yes I'm going to start by explaining a little bit the complexities of installing a networking solution open site usually what it happens is that you have to have one stack environment in order to connect a networking solution to neutron but at the same time networking is kind of the fundamental element that connects all the pieces together so in most of the deployments what happens is that you have to go through an interactive process where first you bring some servers then you bring the open stack controllers you install the network you configure everything and you have kind of multiple dependencies and going through the cycle to have everything up and running what we saw when we started working in juju is that somehow expressing these complex relations within the juju charm and creating a bundle that encompasses the whole thing it creates basically the simplicity that hides all the human steps within the process and it basically prevents errors that was the most powerful thing because when we live in this world that basically you have to install test create continuous integration environments have your development environment replicate exactly the same environment for your production environment or your testing environment basically the juju bundles with all the relations between the open stack and plan grid that was what simplified the whole process internally and with our customers so that was very powerful then compare one of the things that I wanted you to talk to a little bit because I think not only did you invest a lot of time in building those relationships and creating the bundle itself but the bundle then becomes available through our juju store to anybody to reuse simplifying the hard work you've done yes putting in the hands of many others if you think the other thing that in the open stack community is kind of difficult is where do you fetch the packages the information and the software that you want to run so we we work with Canonical to put this in the in the juju charms marketplace and without even us realizing like last week somebody wrote a blog about how much SDN was being used in the while and it was surprising to us to see that there were 98 deployments of plum grid with juju charms and we didn't have to do anything we didn't even know that people were kind of deploying it installing it and trying it just by the fact that it got published in this marketplace that everybody has access to that was a pretty good surprise and one of the hidden benefits I think of this is what we've heard from the marketplace is that it's very difficult to deploy open stack and then adding an SDN made it even harder so if they're doing an evaluation of one SDN compared to the other they really would just invest the time to deploy one version of in essence one open stack cloud with one SDN and by the time they actually had that up and running that was good enough and that was the one they ended up using what we were trying to do in terms of enabling juju charms and autopilot to simplify the deployment of open stack and then simplify the deployment of an SDN with it is actually allowing customers partners to do comparative analysis so when we've talked to carriers for example now they're much more comfortable being able to say well I'll build up one cloud with this SDN I'll build up another cloud with another SDN and do some actual comparative performance testing can you explain a little bit you know the plum grid interface and how you differentiate differentiated plum grid from what you think you do well and what you think the customers like about it yes as you can see in the charm here you have multiple components these three blue components plus the purple one are the plum grid elements and we have what we call the director cluster which would be the controller environment but we have something interesting that call the plum grid edge this is a piece of software that we deploy in each hypervisor that allows us to give things beyond the traditional OBS and the standard SDN solutions the idea here is that plum grid provides an overlay based technology a software SDN solution that was focusing on isolation security and extended capabilities so essentially we plugged into neutron and all your networking needs for an open stack environment from switching routing not floating IPs security groups DNS DHCP basically we provide all these things as a solution in order to that we needed a way to deploy the software in all the compute nodes that would essentially simplify the deployment this is where the charm that we use once installation is done what happens is that you have your open stack now you are in open stack plan in this case we have already created let's say two networks and some instances and a router connected to it so what we do is all the networking technology that happens within this open stack environment is provided by plum grid and I'm just going to go here to the user interface that we have to explain you a couple of context the first thing we do is we focus in creating kind of a secure SDN solution for open stack the reason is because when you have multi-data and clouds you don't want one of your tenants to interfere with the other ones you don't want to have traffic that crosses and you have to have proper isolation so with it is we introduce this concept of a virtual domain a virtual domain is kind of a sandbox from a networking point of view that all the VMs all the compute elements that you connect to that network are going to be based on their identities connected to this virtual domain and in there as you can see basically we have the switch that or the network that was created by open stack with the address management the HEP the router that was created by open stack and the other switch with a set of virtual machines that were jump-started in this case you have here the rules to onboard the VMs so that's kind overall what we do basically we provide a solution from a security point of view scrubbing high availability for open stack the other thing that was very interesting working with canonical going back to the juju charm thing was that what you can see here is this was to deploy open stack deploy open stack and plumb grid but then on top of that you use exactly the same mechanism to deploy applications on top of open stack you may have a spark application you may have a web 3d a web application load balance a load bouncer so what we started discussing with canonical was how does the network player role within these deployments and the ideas that now we are in this demo specifically focusing on open stack I mean all the things that you could deploy in this environment would be on top of open stack but the charms and the bundles that are being created could map to different structures to a few of slides ago we were showing Mark was showing this slide that you could see Amazon open stack Google and the idea is when you create applications and have to burst across clouds how do you do that and this is what we are going we are doing some interesting work jointly with canonical where how do we hide the network complexities of this little green arrows basically the connectors that connect applications when you go to a charm that sense is trying to deploy an application and how do you make that the network is not in your way that you focus on your application you have your bundles and your application bundles and when you deploy them automatically the network underneath has to provide the proper connectivity with the security and isolation properties that your application may need so we are doing some interesting work on those directions right so what I think it's important here is that we've set the stage to simplify the deployment of open stack then the bundle with the SDN then ultimately as Paris said trying to ultimately expand the network and treat the network as just a pool of network nodes if I may John I can show you the the network topology impaired just touch on this this is many of you see this before this is of course horizon and the network topology and being shown in horizon these are a number of workloads that are running in in our cloud plum grid is of course providing the network in this case you can see actually it's a it's a number of containers so these are LXD LXD containers that are running and plum grids providing the networking interfaces for those so it's supporting container-based workloads running in an open stack environment thanks Mark and then well well machine so it's good when some bits of the demo work right so so at the end of the day there's a couple things that we're enabling here that that's very important one is we're bringing this simplification and scale and two thanks and two we're enabling through automation the sort of level of scale that you see in other industries so when we talked to telecom providers they live in a world where one system admin manages hundreds of machines when you go into the Google Airbnb side of the house they're able to manage tens of thousands of machines with an individual all through automation so when you look at AWS they were able to open a data center in Singapore for example with thousands of machines with three individuals right so that's the kind of that's the scale that we're trying to bring to the enterprise into the telecom market let's see so one thing that we've also done and we were thrilled to see this open stack summit is the most recent results of the user survey and this slide shows a little bit of the data what we're really excited to see is that of the major clouds in open stack a thousand users or more the survey showed that we had 65% of all those clouds and overall across all clouds we had 51% but the exciting thing for us is that we also grew 10% over the last half so from our perspective we see us having majority market share and on top of that growing faster than any other part of the other options out there sure so that's a combination of all of the above and if John probably jumps onto slide 10 sorry not knowing where you're going yeah the this gives so we operate in many different areas across the open stack sphere so everything from just being the operating system underneath to people using a bunch of open stack or packages to people using canonical open stack which is our packages that are driven using our reference architecture and tooling to do mass etc an autopilot and then the final option you'll see on here the boot stack so where we are delivering a fully managed on-premise service so it's all of those things some of them will say just be the OS others will be people like bityle or my media and p1 that have got a fully managed service implementation and what what we've seen you know some of our customers and some of our mutual customers they start on one part of the equation and they they go up so where we've been in cases just the operating system we have a dialogue with them where they're starting to look at how to automate and simplify what they're doing or how to leverage what we're doing in the SDN space to simplify the deployment of open stack with the SDN now we are bouncing around a little bit but it's the end of the day so I decided to do that so let me just spend a second here on what we are putting in place around open stack and SDN you saw how we're trying to simplify the deployment and the work we've done with plum grid and the excellent work they did on not only creating the charm and the but but the bundle itself we've created a set of partnerships across all the major SDNs in the marketplace Juniper plum grid Cisco come to mind they're in the process of creating charms a number of these terms are already created what that allows us to do is not only supply choice in terms of people mapping the SDN that's right for them against an open stack that's configured correctly for them but we also allow them to have choice on the hardware on the server on the storage side hypervisor then on top of that what we do is choice by itself isn't any good if you don't have any confidence in it so we use our open stack interoperability lab to do thousands of tests a month to make sure that we mix the configurations of the cloud and map it to different applications BMS the ends of VNFs and make sure that it doesn't break just to give you a sense of the magnitude we have 73 hardware combinations which result in over 175 cloud combinations are they said thousands per month 46,000 deployments and ultimately we have many many functional tests the end result of that is having a degree of confidence that when we have somebody deploy the bundle as you saw with the plum grid charm it'll actually work so we're getting very close to the end and I appreciate everybody taking time and staying with us at the end of the day what we're really trying to do is make open stack easy open stack itself the management of it the deployment of it and ultimately deploying it with SDNs and other applications we recognize that automation itself is not enough you want to be able to create this automate it reuse it share it you know and as pair pointed out you're actually you can put it out in the marketplace so to speak and let other people use it I think that's good for open stack in general the foundation and all of us that are trying to drive adoption lastly choice and simplification go hand in hand for us if you give people choice and they don't have the confidence that it'll actually work that when they take a particular hardware device particular server particular storage deployed in the cloud and add a particular SDN and they don't have confidence that will all hang together we're creating some of the problems that some people have had over the last couple years with all the difficulty of deploying open stack and what we're very confident is with Juju and the autopilot we're taking that complexity out and people are able to deploy clouds quite simply with that I think I'll take a pause at the center stage and I'll check on our cloud deployments going so see how far we've got 30% those of you wish to stay till seven o'clock can see it does anybody have any questions in pair please come up and join us for any questions and yeah please yes so the question was are these charms available if you go to the juju charms as Marcus feverishly typing you can actually go to this website and you'll be able to pull up the charms and the bundle itself so it's not just the individual charms that are available but as we said the work that Plumgrid did for example and creating a bundle with their charm is available for people to use not only to use but make their own version and modify it for their needs and one of the things we're trying to do to simplify open stack is give you know in essence a starting point where you can build on the success and experience of others and not treat everything as a clean sheet of paper exercise so this is the goal of charms really is that we are crowdsourcing knowledge and best practice from as wide a field as possible so we've developed the charms for Plumgrid and for example in conjunction Plumgrid actually took the lead because they know their application better than anybody but over time as we get feedback from users the deploying we will see we want to get input into those right how best to to be able to do these things and so John lead the demo now so there are many charms is about 130 or so charms in in the store that we verified and validated publicly available of course we have many customers that will then create their own branch and we encourage them to then if there's things that they're more broadly applicable to push those as we call backup stream into the charm and make them available for everybody else yes sir we make class with multiple rabbit mq's so and we typically use mirrored cues in them yeah no mirrored cues generally I mean we have customers that do it in multiple different ways our best practice architecture and as part of the autopilot when we deploy it will be deploy typically three rabbits with mirrored cues okay yes what what what would be the advantage of clustering versus mirrored they can be problematic yeah and so again the charms that we've developed anyone from the charm team here no nope so we have a team of charmers that work on the on the on the episode charms a lot of there's two things a lot of advances were made in the kilo cycle with rabbit and how that was configured and secondly we did just prior to the kilo cycle a whole set of scalability testing and so where we launched 168,000 VMs across 560 odd machines and there we had a lot of rabbit problems and so a lot of that config was put back into the challenge to enable us to be able to deliver that kind of scale but if you want to dig into that a little more then we can certainly connect you with one of the architects that will be able to help sort of shape yeah yeah any other questions any questions for pair it's late yeah the the Wi-Fi is a bit slow because we're going through the VPN in Boston yep so this is the canvas of the charm store it's juju charms calm so to your question are they available yes all you have to do is go and you can see them yourself sort them by plum grid open stack juniper open stack etc and you'll see the bundles as well as the individual charms I think we've all had a long enough day so thank you very much for taking time we'll stay around for some questions and greatly appreciate the guys that stayed at the bitter end thank you thank you there are some feedback forms if you want to fill out a feedback form and you can either leave it on your seat or pass it to somebody on the way out you can win some buntu swag thank you thank you