 All right, I guess we can get going My name is Dmitry Ivanov. I'm a manager of technology architecture in the city office and tell us Michael back is a technology architect in the team and between the two of us we will share our experience with This is a private cloud proof of concept with Red Hat Cisco NetApp using the Red Hat OpenStat platform is our cloud manager and using Cisco in NetApp using FlexPod as our infrastructure So quickly quickly on the agenda. So, you know, why we think private cloud? What design criteria and approach we came up with how we selected the technology and the partners Michael will talk about the actual technical implementation of the Of the proof of concept and in the end I'll cover some next up and some and the next steps and some kind of general thoughts on OpenStack and how we can help the community and OpenStack grow So quickly on the background. I will Just roll back about two years ago early in 2013 You know at that time tell us was Obviously, you know already highly virtualized. We were we densified we consolidated We were reaping off all the benefits from from virtualization. Let's put it that way So one area that we weren't, you know, very happy with was the the state of our infrastructure delivery It was fairly slow unpredictable Inconsistent, you know how one of these kind of typical request-based system where you know somebody would You know put in a request we would go and You know a designer will take that they will Go come up with a with a solution And and get a number of deployment teams involved to you know go deploy this thing and The end result as says as is that slow uncreditable Inconsistent, you know all of this so we wanted to change all that right we wanted to Kind of what we Envision at the time would be you know, we wanted to have this, you know a nice portal Services catalog, you know go, you know, select a bunch of Items one or two or three from from the catalog that you need You know you click the select button and you get those Those provisions very very quickly So we kicked off a project. We call it end-to-end automated provisioning and by the way end-to-end We didn't want it, you know just to be the you know just the compute here, but You know the entire scope of the infrastructure services with the networking storage, you know firewalling load balancers all that so We kicked off the project we We selected an open stack based Automation tool set And I'll just roll back not roll back, but you know fast forward Forward rather to kind of almost two years towards the end of 2014 and you know by that time We We achieved some pretty good results. So first of all we learned a lot right we learned a lot about technology but I think was most importantly we learned a lot about our organization in terms of you know the processes and the The organization itself our processes can we realized very quickly in that In that process that you know every team knew what What they were doing they knew how they were interfacing with with other teams But nobody had an end-to-end picture of what that process actually looked like so that was very positive We actually learned that documented that that was That was great So we also we also demonstrated that we can you know we can do automation for some For some very simple use cases At the same time, you know, we we realized that this approach was Kind of we faced quite a few challenges and and limitations So, you know in essence what we're trying to do was to you know transform a legacy I wouldn't call a legacy but a traditional it it environment for delivery of cloud services And what we found was that you know first off it requires is you know fairly complex integration not fairly like very complex integration, you know between our the automation tool set and You know the variety of Management tools that we had in the environment, you know all the ones that I have on the screen IP address management resource reservation capacity planning all of those so That wasn't the only thing though if even if you you know if you manage to You know to do that integration then we have to go and kind of live with it with it is so we we found that it's it is Also very difficult to maintain it because we created, you know a number of interdependencies between between tools that were managed by Different teams they owned by different vendors some of them were you know internally developed So that again was another challenge and even if you did all that You know you haven't done any you know we haven't done any automation yet You know you just kind of built our automation framework now. We had to start building or creating You know a fairly complex workflows on top of that in order to be able to you know to do the actual automation and The reason we have to go through you know all that trouble was If I can only make this thing work if I knew how Yes, it is this bullet bullet point right right here You know it has to do with the fact that the legacy infrastructure Or traditional infrastructure. I should use this probably a better term. It's just not suitable for cloud delivery, right? You know compute is the only tier that is That is you know virtualized or you know software defined whatever we want to call it And what that means is really that the automation framework You know can talk to you directly to an API in our case was the you know the the the v-center API and manipulate Manipulate those compute resources VMs basically Directly that wasn't true for the rest of the infrastructure not for networking storage firewalls You know and all that So in the end of the day, you know, we found You know two things that we can it can work to automate You know the provisioning for some from some simple use cases But you know, it was not going to transform a large complex You know traditional environment into a cloud environment and by the way, there's nothing there's nothing wrong with the with the environment It's it's we're actually very proud of it It is just you know, it's not designed for you know for cloud delivery, right? In this diagram here kind of demonstrates what I just what I just explained so you have the You know you have your silos at the bottom You have different silos have their own configuration tools on top of that you have a bunch of You know management tools that I just mentioned and on top of that you have the automation automation tool set And this integration that I was talking about is you know between the automation tool set and all of these different Different tools that I was I was mentioning so it's very complex on top of that you have the You know the actual the actual workflows Again the complexity You know the complexity and the kind of the issue the the challenges and limitations I mentioned are right here in this You know, it's not really the automation tool set. It doesn't matter which one you you know you select it could be You know You know You pick them right the big cloud UCS director you know different different Distributions of you know open-stack base tools Doesn't matter, you know, you're not gonna kind of the fundamental problem is down here. You're not gonna kind of Solve the fundamental problem the problem with the foundation with you know something in the very kind of the You know the you know the top floor of the building really right so This is the I'm not gonna go through through this this is the this is what this automated the workflows Look like as I mentioned if you're kind of up here, you know, you do that integration now You have to do workflows on top of it This is this is what a workflow looks like for you know just be building a you know building a VM I'm not sure if The room is in the room. I Hi It's his light. I shamelessly stole it. I just wanted to give him credit. Thanks a room Okay, I'll get to that later. So Anyway, something else interesting and that was really the way the most interesting thing that happened late in late in 2014 You know our discussions our conversations with our, you know friends from The development organizations kind of started changing in this way Basically, you know for first time and I'm kind of surprised and we shouldn't have been because we knew that was coming But it changed to a point where they say look, you know, we don't want to don't want it to kind of Spoon feed us infrastructure. We want to give you access to services And we're gonna provision our own infrastructure and we have this specific specific use case and I can't I can't disclose vendor names or product names but the the case was The was this basically they said look it's a very large system The only way we can we can deploy it. We can manage it is through cloud services You just have no out of way we can point at this thing to you know the deploy to one of the existing You know the public hyperskills like AWS or Azure We can put it to an open stack Base private cloud if you have one but the only way we can do it is through cloud services And by the way, we need the full scope. You don't just want to compute. We want to be able to You know Create the private villains public villains. We want to be you know, want to do firewalls load balancers the whole thing And it has to be it has to be self-serve obviously it has to be elastic too, right? we're gonna, you know scale up and scale down and we We So we have two things going right on one hand, you know our kind of current approach with you know, what I call a brownfield trying to Sort of integrate an automation tool set within within the traditional environment And we kind of knew that this wasn't going to get us to you know cloud services anytime soon And we also knew that we wanted you know, we wanted cloud services so basically the You know the requirements were for you know something like this and at best You know what we could do with the you know We just automation in the traditional environment was this which is you know a bunch of kind of pre-canned You know use cases, you know You know services catalog where you know you can go and select and build which was you know Which is pretty good, but it was not it was not this it was not cloud services, right? So the the result of all this was that you know We kind of the only conclusion we could make is that we needed a we need a fresh start We needed to you know start looking at a at a greenfield approach for you know How we built a you know a private cloud for and that's how we kicked off this POC So the first off, you know our design approach. I already talked about a lot of this You know build on a greenfield with dedicated network security, you know the entire infrastructure perfectly Purposely designed for you know for cloud delivery and what this means is You know, I tried to show on this on this diagram here first of all, you know no silos, right? second the the the Cloud OS, you know has to fully control the infrastructure if you look at from the infrastructure perspective the infrastructure has to Give up all control All control to the you know to the cloud OS Very much the same way as if you you know if you deploy a bare metal OS on a you know on any server You don't manage that how do we in any other way? But through the you know through the operating, you know Linux or whatever other bare metal OS you deploy you know say same concept and the third is That any You know any services from that infrastructure that can only be delivered through cloud API So that's the reason I I drew the whole thing circular So there's no corners, you know people can go around and and say hey, you know, I want infrastructure Some out of way, you know the only way it is true, you know cloud services So remember I talked about complicated You know integration with enterprise to set well Don't do it and or at least, you know, don't do it on the on the cloud side do it on the tenant side True through the API. So I'll talk about this a little bit In more detail a little bit later on The rest is just really a definition of a cloud, you know programming interfaces full sets, you know multi-tenant I'm gonna spend a lot of time on that not that the extremely important It's just you know, they kind of self-evident, right? Talked about this so at a very high level This is what you know, we kind of envision the cloud with would look like right there will be, you know, two data centers east and west two availability zones in each of In each of data that these data centers completely independent. So these are basically, you know four separate four separate clouds You know we there with their own a gateway. They'll obviously be a high throughput low latency connectivity between the the availability zones in the In the same data center, you know all connected through You know to our MPLS backbone. This was kind of similar to what This this this large-scale application large-scale service. I mentioned it was kind of Not kind of it, you know was was part of their of their requirement as well. So This next slide is a little bit more is a little bit more interesting because it shows the relationship between between the cloud and the tenants so you know first off They you know as you can see they they they don't share they don't share anything the the only way the tenants And and these tenants by the way, these are the you know traditional Traditional IT environments that we you know would have in the in the organization. So the only way You know the tenants can consume services through the cloud is through the cloud gateway and You know as it's kind of normal you'd have a control plane We have a data plane your data plane would have the you know will do You if you have applications and services that's pan across You know between the tenant and the cloud that will have your file transfer database queries web queries, you know RESTful API's you know all that the the control plane would have your you know would be Your cloud cloud API's and that is really the only way the the tenant can request and services from from the cloud As I mentioned if if you need to do if you do need to do any integration between Enterprise enterprise tool set And the cloud you have to do it on the tenant side, you know, I wouldn't I wouldn't recommend You know doing it on the cloud side. It's no different than the way, you know any tenants any treated by any of the existing You know hyperscale cloud providers, right? They're not if you let's say you want to create something something like a like a VM or a vlan And we want that to be reflected, you know, let's call the configuration item like it's in cmdb terms And it won't that to be reflected. You know cmdb here You're you're not gonna see you know support from your From your provider Other than you know what is available through you know through their common common API's So if you want to do that integration do it here, you know, don't don't clutter your cloud. So You'll notice this You know any V any of the managers here and they're you know a bunch available from And we haven't made any selections, but a number available from you know from companies like like Ericsson Huawei Juniper ALU So the reason I'm mentioning this is just they're just gonna go back to a slide over here You can't plug those with Waiting though. Yeah, okay To this you can only plug it to this right so just to reinforce that You know just a bunch of pre-canned You know catalog items in the services catalog is just not gonna do it. You need the We need the cloud services All right Okay, so now that we we kind of knew what the this cloud is is gonna look like It was time to you know select some technology some partners From partners perspective quite frankly, we you know, we knew at a time that we couldn't do it alone We needed we needed help so we we partnered with them and in Michael and I gonna talk about this so first of all from cloud management perspective Was a very easy You know very easy decision, you know wasn't wasn't even in the consideration that would be you know selecting anything else Open stack is you know widely adopted by the you know communications providers across across the globe tell us is Tell us is no exception. You know, we already committed to open stack You know for any V and SDN we already had some good experience with the open stack based products It is the leading you know initiative open-source initiative for You know cloud management As I said an easy decision. We weren't gonna do it alone In terms of we you know, we were not going to have our you know sort of build our own distribution of open stack in this case we decided to You know partner with red hats. They're already a strong partner of ours for you know OS and virtualization You know, we like their leadership in the open stack community the track record in You know delivering, you know enterprise class products tools You know, we didn't open with an open-source business model So with that I will hand it off to Michael who will cover the rest of our technology choices and the technical implementation of the proof of concept and thanks to the turn So for a compute and network infrastructure partner for our proof of concept or this proof of concept is Cisco We have a wide deployment of Cisco compute and converged infrastructure a number of our data centers and environments We have a proven sense of enterprise reliability and scalability through those systems and we have some familiarity So it just makes a good sense Cisco also has a strong commitment to open stack They have some reference architectures that are directly linked to the flex pod and specifically to the RDO distribution of open stack And you can find those at this link here Our storage infrastructure that we are using for this proof of concept is net app Like Cisco and that tell us how's a variety of net app deployments in our data centers environments Again like Cisco they have a proven flexibility and resiliency from an enterprise bubble We know that they have a strong leadership commitment to open stack. They're one of the first contributors to the open stack code base and Like Cisco, they also have some good reference architectures and tools for us to refer to to deploy And given that we are partnering with net app and Cisco it just makes sense to deploy our open stack to a flex pod There it is In this case, we are using a nexus 9k to fabric intercom X We have a UCS chassis that is Filled with the eight blade servers and we have a net app faz controller with cluster data on tap That cluster data on tap gives us the base things like continuous operations They have that storage virtualization technology that maps really well to open stack You know, they have the net app features You know, the other vendors have this to some de-deplication Net app has a mature center driver and there's the link to the reference architectures They have and they also have a reference architecture that is specific to the flex pod and red hat distribution of open stack These slides will be available Pictures oh, it's selfie you got a selfie. I see I see okay the The net app array was pretty straightforward to set up You know, they have a nice gooey that you can go into and here I just took a screen grab of where we're setting up the storage virtual machines We can apply some of the net app features, you know, the thin provisioning and de-deplication to the volumes that are providing up to open stack The storage virtual machines we do we do, you know, just in this proof of concept that does provide a nice sense of Isolation against other tenants that we may later put on there and that's sort of our The method they were going to use for multi-tenancy when we when we get to that point And then there's also a similar gooey for setting up the UCS chassis One feature that I really like about the UCS chassis and the UCS system for in particular. I mean, I know it's good for You know in general production operations is that they have this idea of the stateless compute Where every blade server is provisioned via a mechanism of a service profile So the service profile has something like I don't know 96 Unique uniquely identified configuration points that you can just apply to a blade when you put it into your chassis And for a proof of concept, that's very nice because you could set up different service profiles and apply them to the blades as your needs change And I show these guis but once you have your environment set up and open stack is running you wouldn't return to these guis You know, it's not something you have to continually return to to keep your environment running You know after the setup all of the provisioning and configuration that you'll be doing will be through open stack The partnership the consulting that we had with red hat was quite useful for us They spent a good deal of time with us. They came in did a planning session where we did all of the sort of hardware layout and Logical network designs. We did everything from setting up which Vlands and subnet we're going to use all the way down to Host names and what nodes are going to go on which server? And then after our planning session they came back and spent Five days with us. I mean, I guess that's a rapid deployment of open stack. I don't know But we spent five days setting up the open stack and we went through and we did a validation that everything was open running and doing what it needed to do And at the end of it, they provided us with a nice little Rebuild recipe that we can you know, we have our own sort of reference design guide now I should mention that we use for our install pack stack I guess the other option would have been to use the rail OSP, but we wanted to do rapid Deployments, so we use puppet and of course that provided us a bunch of modules that we can refer to and tinker with them modify as we go And now that we have our open stack environment up and running we can begin to explore And provide to some of our other internal teams their pox, you know I know that there's a team at Tulsa's that interested in platforms of service So it'll be nice for us to kind of explore paths with them. I Know that our security team is quite interested to see how we're going to Merge and use open stack within the Tulsa's environment You know as I begin to kind of show this environment out to our enterprise network How am I going to ensure that I'm not doing terrible things? We have a mandate to work on NFV and SDN and Then finally what we're going to be working with the I guess it's not in the net apps Mental if I'll show service, but we're going to explore some of the other storage features that are available and I think with that I'll hand it back over to a dimitar to talk about some of the more meta next steps for time Right, right. Thanks Michael All right, so quickly on the quickly on next up So these as you can see there's a there's a lot of work ahead of ahead of us So, you know priority number one will be these these two We got a we got to figure out, you know what our cloud gateways are going to look like remember, you know on this on the slide I talked about before I talked about You know control plane and data plane and you know the firewalls and so there's a lot of work to be done there With our partners from networking security Same Goes for the for the next item the multi-tenant C multi-tenant capability is how we do the tenant isolation how we apply the the security controls within the within the environment and You know service assurance capabilities building our only internal capabilities to be able to manage clouds, you know You know cloud capacity management practice, you know financial model, you know, how are you know, what are you going to do charge back show back, you know how we're going to meet or how we're going to you know do Charging billing, you know when you when you get to that to that point so Some some lessons learns here in the in a forms of you know things that you know from take these to the grain of grain of salt By the way, it this is based on based on our experience. Your environment might be different You know again, but you know, I know already talked about, you know, many of those don't don't try You know if you want to deliver cloud services don't try through Try to transform a traditional IT Environment into that you can you can still do and make sense to the automation in that in that traditional IT environment Just be realistic, you know, what kind of expectations you would we would gonna kind of have from from that effort If you want to do, you know cloud services if the kind of if your environment is just Kind of complex as ours, you know, I would I would just start on them, you know on a green field I already mentioned about the tools Don't try to try the integration on the cloud site if you have to do it, you know, do it on the on the tenant side Don't do it alone, you know, you probably could it's just not It's just not worth the kind of the time and effort You know, we don't have Our experience with you know partners was you know was great You know find partners who are you know motivated to be working with you investing with you similarly with you know within the The internal your your internal organization don't work in isolation, you know find partners find sponsors and supporters on the We tried this, you know, that will do it and they'll come approach. I didn't work for us You know, it's it's much better if you find real these cases that will be, you know, serve serve world by serve world by cloud delivery and now I'm shifting gears here a little bit away from the from the PC So we we're kind of we're coming to these conferences. We're told that hey, this is our this is our chance to Kind of voice your opinion of kind of what do you want? What do you kind of what are you asking for from this? You know for the community kind of which direction is going to go and So is is open stack the next the next game changer. I suddenly I suddenly think so and You know, here's my kind of my criteria You might have a different one, but that's what kind of serves me well to kind of see if something is a something is a big thing if it's a You know, it's gonna it's gonna set a trend for a long time You know first off does it solve a very large problem? Next, you know, is it doesn't have the you know the functionality that is required Does it hide a lot of a lot of complexity compared to the complexities? You know, that's being added You know, do you does it is it high? Does it have high level of usability and what I mean by usability is you know, easy to access easy to deploy easy to use You know, it's stable. It's operationally operationally sound You know, it's nice to be low-cost and again in relative terms You know and and lastly has a wide industry support both from you know tech providers and as well as tech consumers and a Linux and Unix are kind of the prime example of You know for kind of being a game game changer, you know first you know Linux Linux, you know solve the problem and which problem it solved was the you know the problem of basically you know abstracting the Abstracting the The computer right the single computer, you know providing a programmatic programmatic interface To you know to a single to a single computer then you know what Linux that is True kind of reinvented that through open source and lower the beta to entry for you know any for anyone to be able to kind of afford You know in the use computing literally not just organizations, but but even individuals right and Open stack kind of could be Could be on the virtue of doing doing the same for for the entire infrastructure stack, right? You know first AWS Azure Kind of solve the problem Invented the software data center or you know cloud OS data center OS whatever you want to call it You know so now open stack through open sources I think it's kind of poised to lower that beta to entry again at that level at that scope to make that easy you know of all available affordable to to to everyone and You know software defined we talked a lot about software defined You know, I just I just threw this out there I hope you know, nobody's gonna be offended because this is this is really a basic But I kind of speaking with with the people for the last of the last couple of days You know, I don't think it's it's really Not so much that it's not understood, but I think it's it's kind of watered down a little bit and it really is You know, it really is this simple You know similar to the way Linux abstracts the you know computer hardware so you can you know You can manage your processes files Network connections users and so forth and so on through a single programmatic interface doesn't matter what hardware you have You know, which which the CPU using you know, which vendor You know build it and all that open stack is Is going to do the same for you know the end-to-day end-to-end infrastructure in the entire Data center, it's not going to do it. It's already It's already doing it right so you can You can manage your VMs you manage storage, you know again programmatically through an open standard A standard interface so this is extremely the reason I put it out there because this is You know in my mind, this is you know extremely Extremely powerful, you know when you I think was the gentleman from PayPal yesterday the keynote who said the only thing that is between my Infrastructure my developers, you know is the is our API's right and if you give a bunch of You know solid API's to You know to a bunch of developers, you know, wonderful things are going to happen And this is really kind of why this is this is as powerful as it is All right, so Can open stack do this and do we care and so what is this it's a You know, it's an adoption curve You know kind of which is typical for these game changers You know and as you can see There's there's a lot of work that needs to happen in terms of adding, you know function functionality and usability maturity if you will You know for a relatively kind of low level of adoption and at some point You know over here something happens all of a sudden it it kind of it kind of explodes Everybody starts doing it. Everybody starts starts using it and And You know and so why do we care about this right if open stack is going to go to this point if you know if we're this conference I believe you know, and I think all of us believe that The only way well it can it won't be maybe the only way but Moving forward infrastructure is going to be delivered in a way that it can be It can be consumed through these programmatic Interfaces through, you know through cloud services and this and this line here that is the divider You know between the the halves on this side and the have nots on this side, right and and over here You know, you would have You know Amazon and Microsoft and then VMware perhaps and over here, you know, it's the rest of us behind this You know open stack community Trying to you know trying to catch up and this is why it matters You know for us to kind of as a community to help to help, you know push open stack you know through this Through this point so it really becomes you know functional usable, you know stable operational ready And it's you know, it's easy for you know for us to Not only consume the resources. It's already It's already easy enough, but to also Be able to you know to manage it properly and deploy it easily so and we You know, how can we help right we all have different roles in the in the community and I will just we're running out of time So I'll just touch base on the You know the developers here You know more functionality is great But we need to we really need to focus on usability make it make it easy to use easy to deploy You know easy to manage at this point if you ask me if I even if I have to choose between between the two At this point I would I would I would use more Usability more accessibility For the technology providers are very you know a very similar technology providers They you know they had a lot of these developers here. So you guys can also help with you know driving this You know open stack towards being being more stable more mature You know easy to deploy easy to manage You know for service providers Internal external doesn't matter. You know if you haven't started it get going You know and if you if you already have started just just don't give up. It's hard. It's difficult But you know keep driving, you know voice your ass conferences like this, you know talk to your You know talk to your vendors talk to your providers make sure that they They hear you and for infrastructure professional software developers if you guys already at the conference, you know You figured it out. You already doing it, you know keep doing it and This is the this is the end of our talk Tell us the future is friendly Thank you very much. I don't know if you have a minute for you know question or two If we don't one one question If we don't this hour contacts Michael and I will be happy to you know talk to you guys Offline share more experience After the setup during the setup we we had some struggles, but Yep, we now have through our deployment a great availability I think I think that the the post build experience is quite good It's the getting there there you really we really did lean heavily on our redhead expertise to get that going And again, you know, we do I'm sure if you go to the Red Hat site or contact your Red Hat Consultancy they can provide just more deeper detail and and you know, like say they did provide us with a stack of documentation of how we got that to go Attempting to do it myself. I failed so There's that. All right. Thank you. All right. Thanks a lot