 folks get started in a few minutes couple of minutes here everybody my name is Kamesh Pamaraju I'm a product manager with Dell today I'll be moderating a session with these esteemed guests who know all about OpenStack distributions so the goal today is for us to understand if how and why OpenStack distributions matter for OpenStack for its adoption for future innovation that's kind of the the key goal for today's today's session so with that what I'll do is sort of set the stage for some of the dimensions of distributions as I see them and have the panelists weigh in as we as we go forward here so what is a distribution is OpenStack project a distribution OpenStack project is a set of projects that our community developed but they are not a distribution because if you're a customer and you want a maintenance patch or a security fix or an update or just support plain old support you can go call any of those 400 developers you got to have one entity that you can call and say hey I have a problem fix this for me install deploy fix maintain patch those are all the things that matter when you are actually dealing with a real open stock deployment or an open stack production environment that's where distros come in so we've seen a lot of this in the Linux world there are lots of distributions we have Red Hat Sosa and Canonical here today who have been leaders in the Linux distribution world and now they're entering the the world of OpenStack distribution so so I just want to lay out a few dimensions so a distribution can be paid a commercial distribution and a non commercial distribution a distribution could be targeted towards enterprise users or just plain you know end users that are not enterprise a distribution could be not all the projects with an open stack it could be just some parts of open stack distribution could be some pieces of it replaced by the vendor with proprietary components and lots of different dimensions so we want to find out today from the panelists about what they're doing with their distributions Canonical has it has a had had distribution for a long time ever since long time right and then Sosa came up with their open stack distribution recently Red Hat has announced a preview of their distribution Dell has a point of view about distributions and in some ways companies like Piston, Cloud Scaling and we have Morph Labs here Nabula also can be thought of as distribution so we have Morph Labs here as well who can then provide their input or their point of view of what they think was a distribution so so with that sort of background I'm gonna quickly go through and have the panelists introduce themselves starting with Nick over there on that end and then I'll open up with some questions and it's an open interactive session so feel free to ask questions along the way so I'm Nick Barsett Ubuntu Cloud Product Manager Joseph George Director of Product Strategy at Dell similarly Dell's been involved in the open stack projects since day one and I'm actually quite a few familiar faces there in the crown Christopher Ado from Morph Labs I'm director of technology operations for the company and we sell a open stack appliance Perry Myers engineering manager for cloud infrastructure at Red Hat and we got involved with open stack about a year ago and are now just getting into the distribution for Red Hat. I'm Pete Chadwick from Sousa. As you can see we have some real practitioners in the field who have been involved with open stack for more than a year in some cases so let me kick this off by asking the first question why do distributions matter for open stack? Sure so I look at a distribution and I use this term loosely it's because open stack is certainly not a museum piece yet but a distribution is a little bit like a museum where you have a curator that decides which pieces and parts make the most sense just as Sousa and Red Hat have been delivering enterprise Linux for a long time there are differences between Sousa and Red Hat because we see different different things we're trying to solve so it's our responsibility to sort of look at the world of agree with everything you said there I'll also highlight that one of the things we do as distributions is make things more of a solution rather than just a piece of component parts right so you can think of the upstream open stack projects as you know building blocks and what we want to do is give something that's a little more turnkey for our customers and it's not just about providing additional software for example like crowbar but also about providing tuning performance working with vendors to do certification and also the amount of testing that we put into the distribution and then maintaining things as you go through the lifecycle because customers are not going to want to update you know every on the six-month cadence that we have with the upstream right now so being able to support something for an extended period of time and deal with things like security updates and so forth so my perspective is really I agree with those guys completely and the distribution should be a stable snapshot of a project and with open stack it's so active and and being developed so quickly and moving so quickly there's a real version of open stack is so which tags of each individual components are going into so for us there's it's really important and we use Ubuntu and we count on their repositories it's in some sense what you guys are doing more flabs it's not like a traditional distribution you're taking an existing distribution and then creating a specialized solution for your customers exactly which is which can be thought of as a solution right right I mean my solution exactly and it's our solution is meant to be consumed really as an appliance so we're on the hook for making sure that open stack works properly and in order for us to do that it's if we go at our own and try and roll our own build of open stack every single time and give us a really stable base on which to build everything else we offer on top Dell perspective obviously everybody on this stage is a partner of Dell's and our goal has always been to deliver solutions to customers and as you can imagine we have a wide variety of customers that we try to service on a number of different technologies and so we we feel that what distributions bring the tables out actually really critical for customers and their expectations of support validation etc when you look at what Dell has brought to brought to market the the Dell open stack powered cloud solution made up of an RA and Dell's crowbar software that we developed plus the you know software from various partners and then services on top of that you know we're we're designing a solution that customers can get drop in their environment start you know finding value off you know just recently announced today we announced a relationship with Morph labs in delivering their solution because as I mentioned what Morph is doing in this space Morph labs is doing the space is actually servicing a specific part of our customer segment as well so we view distributions as pretty critical to our customer base and what they find as important in a cloud solution add to everything that has been saved because I fully agree on everything but I think there is there is two fold the fold where we distribute something to people and when we do so we have to make choices open stack is a project which is the world of choice you start deploying open stack and then you already have oh am I going to use rabbit qubit or zero mq am I going to be using KVM Xen IPV etc etc so one of our role is to test a subset of this make choices for you to make your life more simple and then package it and deliver it in a way that you can consume it and deploy it as easily as possible but there is also another fold in it it's our involvement with the open stack community each of us here are here because not only do we consume up and stack but we we also help shape the future of open stack because we and sometimes building the solutions that the open stack community had not thought of to deliver the solution that our customer needs great thank you so so there's a there's an interesting set of thoughts about open stack being the Linux of the cloud and I think it was Martin Mikos who famously said at one of the conferences it's not gonna be Linux of the cloud it'll become the Unix of the cloud so in essence what he was saying was the danger is that open stack can become fragmented right everybody has their own flavor everybody starts flocking forking things off so I wanted to get some sense from the panelists today is open stack the Linux of the cloud or is it gonna become Unix of the cloud what are your thoughts customer that made wrong choices at the very beginning of open stack didn't understand the philosophy of open source and deployed open stack the wrong way apart from these guys I think nobody here are making the same mistake when we want something in open stack it either talks to the official API or it's inside of open stack developed with the community as a full project so the fragmentation is something that Martin Mikos may like because it helps him sell his solution but it doesn't exist so that happens only if somebody decides to fork open stack and create their own distribution and not go back and contribute back to the core right so what do you what are you guys seeing in that space well I just like to add one thing on there I mean there's the notion of core projects and I think we're all consistent with the idea of contributing back to the community and in fact working first with the community and pulling back into the the product right but there is going to be I think a little bit of divergence when we talk about things that layer on top of open stack like we've got different management solutions we've got different monitoring solutions and so forth and so you might see a little bit of fragmentation there and but at the same time there are I'm not saying that's a sorry bad thing but but there are other projects that are you know coming up in the works right things that are heading towards incubation and so I think there's opportunities for us to at least have the same core infrastructures for those things even if our UIs for example are different because that's kind of cosmetic as long as the as long as the foundation is healthy and and we keep having you know regular design summits like this and keeping the community together it won't fragment because it's it's one group of people like Nick said building it and you know there's it's a great foundation where the people who do want to do their own things like projects that communicate through the core API so there's the ability for people to kind of go off on their own and add feature sets but always going to be open stack I don't think if there's ever going to be some and I think just just to address sort of the fundamental question is distribution does not equal for a white space if you will so there's there's an opportunity for you to ask questions feel free I have one question which is you guys have distributions of your own can you spend quick one minute describing why your distribution is different why it's important what are you trying to do what problems are you trying to solve yeah go ahead please go absolutely it's not there is going to be there is the open stack API is already there it's made made to evolve but it's made so that your management solution can know that it has evolved there is versioning inside of the API the appropriate recomponent or a fork open stack you're going to be enhancing open stack the one thing I'll add is that just like with Linux and the base distributions we're talking about that diversity actually helps the ecosystem right so it's not a bad thing that there's you know Susan canonical and red hat all together because we're all independently helping and we're all helping to drive different facets of it other questions yes there are tools that make that relatively straightforward to deal what you say it is true today the way you describe an image that is running on top of open stack is the same exactly the same regardless of the distribution of open stack so that you're using the same API so you can very well run it somewhere else it will run the same way but there there is one difference though which is each distribution focuses on maybe a different set of hypervisors and there are image differences be going from one hypervisor to another and like you said there are tools that you can use to convert an image from one hypervisor to another so when you talk about you know workloads let's say you know it you have an application and maybe it only runs on rel doesn't happen around on susa that does you can take that same image even though it's a rel image and run it on top of susa's cloud there's no reason you can't do that so that's one way you can transition the workloads but again if you're going from one hypervisor to another there's a little bit of a manual step there but there are a lot of tools to help automate that so that that's what I wanted to get to my next section about so what is that rest so this you know susa let's start with you what how do you describe your distro open stack distro how is it different what's unique about it so I think first of all as we talked about at the beginning you know it's built upon open stack it's built upon open stack sx we've we make sure zen was supportable at the same levels KVM and sx so the red hat open stack distribution is I would say based pretty much vanilla from upstream so very similar what you're talking about you know we basically take the upstream releases package it harden it and go through fedora first so when you think about distributions you can kind of think about there's the and you mentioned this earlier there's the product distribution but then there's also free open source distributions and so we treat fedora as the upstream that we pull into both rel and into open stack what we've specifically done and again everything we do gets pushed up stream but we focused on things like KVM as the hypervisor making sure that works really well for us with our versions of KVM making sure it ties really well to Libvert to help support that and you know some of the other things we did had to do with like AMQ P messaging layer right originally rabid MQ was the only supported one so we pulled that out abstracted it and make QPID work so I mean there's a lot of other kind of smaller things that we did but fundamentally everything that we did was upstream pull it in and now what we're focusing on is making easier for use for the enterprise right so making it easier to deploy install and monitor you know our initial release is going to be pretty vanilla compared to upstream Folsom but as we mature things we'll add in those additional monitoring and deployment capabilities so are you guys focused on the enterprise particularly I mean both Susan Red Hat yeah or are you looking at telcos and service providers and such we consider those part of the enterprise so our distribution is fundamentally different because we are selling it to customers as an appliance so they don't need to know how to set up open stack they don't even need to know about the internals if they don't want to we're based on Ubuntu we we use puppet after after we deploy for all the configuration management but essentially what the customer gets is a running supported open stack environment that they can access through our UI which is one of the other big differentiating factors we're not using horizon we developed our own UI API so we gave back to the fog community in hopes of bringing more Ruby developers into the open stack fold and we fog API and our our UI talks to open stack through the fog API and customers have access to directly the open stack API so if they if they if they are you know fairly technically literate and they want to manage their open stack environment via the API they have full access to that and if they are just trying to get their feet wet in cloud we've really low you know give them a really easy use web interface to targeting a specific set of market segment or we're targeting specifically people who could use cloud okay who could use cloud yeah that's it you know if a marketing guy said that you would destroy him so yeah the Dell perspective is you know very unique in addition to the partner offering that we sell with with art the distribution and then we know we could get into the various colors of what the word distribution and packaging means but you know we deliver an acrobar ISO that we actually make available that includes right now with the Ubuntu bits that we deploy open stack on but again it to us it's a very similar thing right and we are focused right now in our solution on the Dell open stacks a more sophisticated user one that does want the flexibility and the power that comes with an open source type product but then as we look into other places and we start looking at getting into other markets it's gonna be and I'm gonna suggest that you start with Nick on the next question because you seem to be favoring that side of the line there yeah our specificity is that we are targeting users that needs to deploy open stack at very large scale is weird enough so that it can deploy the cloud and deploy on the cloud and this gives us a very nice way to make sure that we can automate all the testing to ensure that the upstream bits that we receive are always functioning on a continuous basis and we've invested in making these continuing continuous integration effort the quality and the enterprise readiness of our open stack distribute test also that we can still deploy on open stack with the same result measuring how that say there is another role that we think we have that is very important trust ecosystem trusted ecosystem because as good as we are is something where you're making sure that this right partner works correctly with you is something very important the same thing happened for storage maybe you want very fast IOPS maybe you want very cheap storage which partnership you go to this is what we work on this is what we want to qualify for you great segue into my next question which is one of the things I've been following open stack for the last two years and this time around I think we are at a crucial inflection point we're seeing a separate parallel user track which never existed before and we're also seeing lots of interests amongst users to start moving their workloads into quote-unquote production so my next question is around you know how what are you doing with your distributions or the way you're coming to market around enabling customers to move into quote-unquote production what does that mean for you his main concern going to production is not going to be the cloud itself the list of applications that he expects to be running on this cloud or the list of developers that he expects to a public cloud provider his expectation is how quickly can I be on my customers so going to production is a very subtle mix of needs what do we support how can it be implemented how can it be integrated with the existing backbone under it's the provider backbone for selling and billing stuff or it is the identity back therefore I don't think there is a generic case of going into production every time it's something a little bit specific that needs to take into account the DNA of the enterprise so so what about things like high availability and fault tolerance is that a requirement right so there's a base set of things that don't exist today there's no HA in core open stack but there is no there is nothing that prevents from building an HA open stack inside of open stack in fact there is there was two component that were causing massive edX this has recently been reduced to one so I mean the complexity of putting open stack in HA is something that a good sysadmin should be able to handle very easily that's like saying that Linux has no HA I mean there's components that provide HA and you integrate them with open stack so things like pacemaker for example could be used to do that and then you have to look at HA both from the the guest perspective and also from the underlying service perspective too so my question is related to distributions whether or not distributions themselves provide those kind of HA capabilities out of the box is that something you guys are thinking about this is something that we are more than thinking about this is something that we are the configuration option I mean that's what it comes down to HA for open stack is just configuration our goal is that with juju you deploy and you do juju adjunct and your HA just say my perspective on it is similar but you know we kind of treated more in a layered fashion so there's one aspect of HA and open stack is the fact that the services are all horizontally scalable right so if you want high availability you make sure you have low balancers you make sure you horizontally scale your services another aspect is the you know automated health check and restart of services and for that you know we have solutions that red hat red hat high availability software and you can pair those two together so we don't look at that as being part of the open stack distribution we look at that as being part of the overall solution so if you want that you get those two components you you know we'll have white papers that will show how to operate those together and that will provide this to us a driven type of thing yeah okay if I could just add to the original question I mean the customers that we see we're looking for exactly the same thing that they look for when they do a production deployment Linux because at the end of the day when you actually launch the workload it's running on on Slez with Xander KVM it's running on rel with Xander KVM the open stack is actually not part of the execution stack it's part of the management stack so customers are looking for certification on hardware they're looking for certified applications that they can migrate into the cloud and those are all things that are that are inherently part of enterprise Linux as well yeah and to add on on to that I think this is really good discussion when we when we take our perspective and kind of where Nick was going and these are that these are the exact conversations that we want to have with our customers that were actually our basis which is you know when you actually go from a proof-of-concept which you know there's certain infrastructure that you need in place there's certain expectations around the software that you're putting out I think that customers it's it's a that it's a stop at POC and then a lot of our customers are starting with POC's and buying POC units and so it's actually really an interesting evolution it it also ties back to where where open stack is right in its own evolution and as you know as you indicated this conference a great indicator of of where open stack is growing and as we start seeing more and more of these types of very strenuous requirements coming in I think we start finding our finding ourselves in a production environment up the floor for questions in a second I have one question one last question for the panel so in terms of where you see this going in the future so sort of a paint-of-visionary picture of where you see each of you guys see from a from your company perspective next three to five years where do you see open stack going where do you see your distributions going one minute so that's a great question and I can be sure that I will be wrong in my forward-looking statement but clearly I don't see any reason why there is less and less reason to run on straight bare metal I'm saying straight because you can therefore I think this this will really become our common kernel and we'll still be delivering the real and we'll still be delivering kernel inside but every operation that your IS will have to do except for maintaining the bare metal infrastructure to be virtualized by the way I think there's a great foreshadowing to the presidential debate tonight by the way giving us one minute this would be interesting to see what's going to happen yeah but essentially three to five years you know again to go back to something I said earlier you know the Dell customer base is varied and there's a unique set of needs that come with a bunch of different markets and where we were right where I see this going in three to five years and in the space that we that commission I are a part of we are also the team that's working on Hadoop solutions obviously what we're developing with crowbar we've developed an open source community so what we're seeing is I believe is more and more appetite and interest and adoption in open source in general and I think that will help drive where OpenStack is going to go put some bold statements out there but I'll be recorded by the way yeah I know that are gonna be recorded I gotta be shown right in next year and when we're all wrong yeah but I think what's what's gonna be happening is we're going to see more and more the enterprise start adopting a little bit more that being said there I believe there's going to be a market for existing cloud technologies as well I believe there's going to be market for all of these things OpenStack offers us a venue to kind of a new space that customers that have the needs that are that are met by OpenStack will be with what Nick said and I think OpenStack is going to become the de facto standard when it comes to delivering infrastructure as a service and like Nick was saying people will stop thinking so much about hardware there will still be people whose jobs it is to maintain that hardware but the end users are really not going to worry about the hardware is going to become more and more deeply extracted the abstracted the networking as well and I think OpenStack is going to be the technology that brings that to the people and kind of takes everything else out so there will be there will always be competitors and smaller smaller people doing the cloud stuff but I think OpenStack is going to become the kind of the go-to standard so I'll go on in a limb and make the statement that none of you guys want to make OpenStack will be the only cloud in five years we can all hope right now I'm sure that it'll take a little bit longer for that maybe six seven years so I agree fundamental with what you guys are saying and I like the analogy that you brought up earlier of OpenStack being like Linux and I think just like you know speaking specifically to Red Hat OpenStack you know initially we have a huge amount of churn we have a lot of features being added and just like we had with Red Hat Enterprise Linux over ten years ago the support model has evolved right so initially you have people that are on a very short support cadence and so I think three to five years from now what we're going to see is a stabilization and you know maturity of OpenStack based clouds so that you know customers can stay on something for a prolonged period of time so eventually you get to the point where you're actually thinking about hey I want to be on a particular release of OpenStack for five years now we need to take a look at today with Folsom does anyone think that you could be on Folsom five years from now probably not but we want to get to that point and that's what we're all driving towards so I think obviously I'm not going to say that in three to five years OpenStack is not going to be successful because so so I guess my my vision three to five years you're all if you're an enterprise and you're running have OpenStack set up and send off a workload to run a Hyper-V set off workload to run everything through that single pane of glass I mean that's and 70% of your workloads are going to be running a 30% that will still be running using existing verification management technologies to OpenStack become the the Linux of the cloud the question is who becomes the CPM of the cloud. Alright with that we'll open it up for the questions on the floor questions yes yeah go ahead so I think it's just like Linux today I mean you'll always be able to migrate if you're using a specific set of tools that's in one distribution and it's not a different distribution you're gonna have to you're gonna have to migrate at least to the tool set level but from an API level you should be fine. Yeah I would just say that the whole point of having a core OpenStack set of API is that that allows you to not have vendor lock-in at that level and if you have lock-in at the tools that's a different situation but I think that as long as we're all focused on working upstream first so if you're gonna work with you know Canonical to get a feature upstream it's going upstream and so therefore Red Hat will also benefit from that same feature. I know this oversimplifies OpenStack but a reasonable analogy is think about Apache if you're gonna run Apache on Red Hat or Apache on Ubuntu if you're running Apache 2.2 you're running Apache and and your HTML and CSS and everything behind that is is gonna be you can migrate it from one Apache under any distribution you want to the other. Yeah and I'll just briefly say right before I hand it over to Canonical here I think it's within our powers a community to help influence this as well right that's that's that's what's unique about what we're doing here in this in this moment is you know there are decisions that all the great partners on this stage are making but we as a community have a voice to actually let the so yeah I don't see any reason why when we are talking about the same version of OpenStack there shouldn't be a in fact I'm quite sure when these iPrevisors are going to be running different OS's and it's one or two persons answer that that'd be good. In production it's in the rounds of 50 and saying some of these numbers but yeah we have we actually have a variety of customers in all these different stages of the marketing team hasn't ever say a number. You can say any number. We work with a lot of service providers and so for the service providers they white label what what we build and support so they've also deployed it and if the if the heart of the question is how many people are using OpenStack it's it's the number is growing every day and it's I can't a full product out yet so we don't have customers in production but we are working with a handful of Lighthouse customers we're calling them to get them to production quickly so that when we do have a GA release of the Red Hat Open Stack product that they'll be ready to go out the door I'm also not in marketing so I'm not gonna throw a number out because my marketing guys will hit me over the head concept Lighthouse customer kinds of kinds of activities going on. We're just about out of time so thank you very much panel members for your contributions