 Good afternoon. Welcome to Cisco's premiere sponsored session this afternoon session number four for us and this is day one of what has turned out to be a very exciting summit so far I want to introduce very quickly our speakers for our premiere session Lou Tucker our CTO and VP and VP He'll be joined with joined by Ken Owens also CTO and the clouds Cloud services group So with that I give you Lou and Ken and OpenStack changing the face of service delivery Great. Great. Thank you. Thank you very much I think what we want to do with this session today is really talk about what we're seeing in this kind of changing landscape and particularly how this affects The what we're calling the face of service delivery how services now are going to be delivered to users On cloud platforms, and I think this is a very exciting time I think you've seen this theme repeat itself in a lot of the other sessions that are going on here in the summit this week and in fact, I think it's I sort of start this this line of thinking by looking at the big disruptions that are taking place in our industry and what's making those disruptions possible right now, for example, if you look at Uber it's disrupting transportation Taxi systems Airbnb disrupting a hospitality and the hotel markets Netflix disrupting video delivery and Amazon disrupting traditional retail These are tech these are new disruptions that are happening because of the internet and because of cloud computing These are some of these very New companies got going very very quickly and are serving a worldwide market Because of the cloud and because of internet delivery that kind of disruption I think is important for us all to understand because in many ways the businesses that we may be a part of are going to be facing new competition from a lot of these disruptors and so we have to be able to disrupt ourselves and so we have to think about What is the purpose of cloud computing because originally from several years ago when Amazon really started getting going I actually was at a startup and using a lot of Amazon services And it was the greatest experience of my life in terms of being able to very quickly develop new applications and services And I could scale them up and down and all of a sudden we started talking about this dynamic API driven way to get resources And this was fundamentally changing the economics of computing because I didn't have to go out and rack and stack servers I could use somebody else's and through API's I could get access to them Run them for a couple hours and turn them off and therefore very much more accurately tracked sort of the demand curve and save an awful lot of money and so the first thing that people started thinking about when cloud computing was a fact that it was an Economic change that you could really be much more efficient But in fact, I think as time has progressed we've realized that becomes secondary becomes very secondary to the dynamic nature nature of cloud computing and that's when we've moved from this physical domain into the virtual domain So now we no longer think as much about how many servers we're running or how many racks and everything else Unless you have to be a service provider if you're an application or a service delivery Application you're thinking much more about virtual machines and about containers and this whole new world where these things become ephemeral You can spin them up. You can spin them down and very quickly deliver new services and Some people also talk about this and the age of kind of software-defined data centers Even traditional companies such as Bank of America is talking about embracing a software-defined data center What they mean by that is that they want to be able to look at their data center and use software to manage it instead of having System operators go out or network engineers and cable things and change everything and logging into machines and doing this They want to be able to do this through software that gives them faster time to market can dramatically lower their operating costs And they're looking then to have open systems behind this because they don't want to be locked into a single vendor for this and Services now they want to be able to span both Services they're running in their data centers and they're running perhaps somebody else's cloud and a kind of a service provider So in this model and wait one way to think about this is in fact Configuration which used to be out of a runbook you would you know have a system operator a login to some server and make a change Document the change go over to another system record that change So you could have compliance and everything else and now we've become to these kinds descriptive models expressed as code This means that you can repeat that operation again and again and again and know that hasn't changed and this is a much bet Much better way and this is the heart I think of all of that we're everything that we're seeing and changing service delivery because now we have a way to quantify Exactly what composes these services and how they're supposed to operate? So software-driven infrastructure now is really possible be based on policy So this is what IT organizations really want to do. They're trying to enforce a certain usage model a certain way of delivering a service Based upon policy rather than the configuration that they might have might have Might have applied so there's been a lot of talk about in particularly coming from Cisco where we're looking at things such as ACI Which is a way to express network policy Rather than what are what are how you want to be able to configure each one of these network systems so that you can talk About policies around this application. Is it internet facing or is it internally facing? Is it accessible from the internet should there be any traffic between this point this server and that server if there Should never be traffic between that that would be a violation of policy So you want to be able to express that and codify that as policy So there's a lot of work being done now an open source an open stack in other areas around this kind of group-based policy This affects very much the service delivery because now you can talk about very large-scale services and expressed as policies That you want to see happen The other thing I think that's happening is is quite interesting is that as a community We're getting together and deciding that it's better for us to work together in open source communities such as open Stack then competing for proprietary solutions our customers are getting too smart Our customers are demanding open systems our customers are demanding that we work together with each other And so in the past we've done that through consortia We've done that through you know getting together in standards bodies And now they're really asking for us to get together and do it through code So in something like open stack where Cisco has been involved since early 2011 We've gotten together and in fact the early neutron networking service that we got together We got together with 14 other companies many of which were our fierce competitors in the market And we decided we needed a network service in open stack to be alongside of the compute service and the storage service And so we work together and evolving neutron to what it is today to be able to be this kind of service that that is vendor neutral The customers can actually use get open stack from many different places And they have an API that they can count on and now the vendors have to compete on the implementation The vendors have to compete on the services they provide around that and the support they provide on that But the actual implement the actual API's we're not trapping our customers into that And so this kind of independence from vendors is something very important What that's also allowed is that now we have a large number of services that are targeting a much larger environment We know the the giants in cloud computing today via Google, Amazon, Microsoft have their own clouds We're looking for a counterbalance to that in this open community where there can be many many more there can be hundreds or Thousands of other cloud providers and and clouds that are being run inside of a company's data center. In fact, that was the original impetus for developing open stack was to develop the software to allow anybody to build their own cloud They can build it in their own data center They could build it and they could become offer their own cloud services So there's a lot of opportunities that being created because we've gotten together around this community-based software And in fact that we've seen and it's interesting if you go to the open stack org site we have a number of case studies and I really urge you to take a look at those because now there's a wide variety of industries and use cases that are being developed in OpenStack and you can find others that might be like your own business and why how they're Adapting open stack for their use so we're seeing things not just from sort of startups and Service providers, but also come big companies such as PayPal or Best Buy and even media companies such as Comcast or financial Services companies in Bloomberg. All of these are using OpenStack for different purposes So this has been actually very rewarding to see the number of different applications that are being developed on top of OpenStack Previously actually we just had a conversation about one of those new emerging verticals. I think which is around NFV Because we're seeing now service providers looking at a cloud platform Which has open API's and saying what we want to be able to do is start to virtualize our network services themselves And I think this is happening because there's another kind of perfect storm that that's developing here Several years ago. We started talking about SDN software-defined networking and and things such as open flow and now we're seeing that was happening at the same time as Cloud computing is happening and those things now are coming together and firming this kind of perfect storm where cloud computing and Software-defined networking network function virtualization is coming together at the same time and creating a whole swirl of activity Around this area as service providers are seeing here's a new way instead of providing You know different appliances that they would go to market with or buy from a vendor They can start defining these things again as software services get a speed to market and get much more Advantages out of the rest of their infrastructure So it was not surprising then in OpenStack itself when we defined a network service that we Defined it such that it would work with many many different pieces of hardware from different vendors And so that's why there's been a large number of plugins and drivers underneath neutron from most of the leading companies here Also outside of OpenStack. There have been other open source initiatives around open daylight This is where the community is getting together and again defining a software-based controller Whereby this can now be used with OpenStack because this is now a software way to define a lot of the network functionality that you want in the control plane directing directing the network to the future So this is this is a Picture that actually a number of us in the in on the foundation actually been working on there's a working group around Essentially OpenStack for the telco industry and this is where you can see a large number of different devices that have been developed Which have been hardened appliances to perform different network functions that are now being replaced by software And each one of these things are now becoming software that creates Requirements on the OpenStack platform for we want this to be carrier grade We want to be able to provide isolation of the resources. We want to be able to have non interference of those resources We need it to be much more real-time and all of these things now are creating Additional requirements on the OpenStack platform that again across the community. We're working together to bring those things about Interesting enough that we see the European standards bodies That's the then got into the picture here and started talking about publishing a spec and here they've defined the architecture For how they want to see NFV working in terms of that we have an orchestration manager We've got a virtual resource manager and other things in OpenStack plays a fundamental piece of this puzzle In fact another group has been started called OpenNFV And this is a group that is developing reference implementations based on OpenStack with open daylight of this kind of an architecture So in this architecture we can see we've totally replaced those hardware network functions with with virtualized services Those virtualized services now have to be carrier grade they have to be resilient They have to be able to scale and but this means that they can be much more rapidly deployed than having a truck roll bringing in new Equipment or trying to expand the number number of network services you can Previously have been done by hardware now you can rapidly bring those up as virtual machines And so this whole new model has coming about we're software as Mark Andreessen sort of said is eating hardware replacing it in a much more agile manner at the National Association of Broadcasters show In Las Vegas several months ago. We're beginning to see this happen now in another industry So it's not just in traditional web applications We saw in cloud computing on the kind of web services We had seen not just what we've seen then in the telco market with NFV now We're seeing it happen in the media creation and distribution And I think this is the one that's next in line and over the next couple of years We'll start to see a lot of activity in this space For example, if you look at the media and what's going on in this world today, there's a huge number of competitive pressures We're seeing over-the-top delivery of video services. We're seeing a lot of consolidation in the industry. We're seeing multi-screen Implementation taking place. We're getting our video and our and our things on many different devices So I want to be able to watching something on TV from a cable box I want to be able to get it on my phone and be able to get it wherever I am in the world All of these things are creating tremendous pressures and tremendous need for these industries to adapt and therefore They are turning to cloud computing for this purpose as well In fact traditionally what we've been seeing has been moving from video-based delivery services and these big switches With a serial digital interface Turning to now IP networks and so this is a big change because traditionally IP was not very good at this You needed to be able to have the delivery of video without dropping any frames You need to be able to have quality of service You need to be able to have dedicated bandwidth on all of these things in particularly around broadcast and ingest We're seeing this change happen now as we're seeing this move to an IP network and Moving it to the cloud So whereas previously we had these dedicated hardware switches now We're using general purpose switches and general purpose servers But we're now controlling them through software through controllers such as open daylight And this is I think again this kind of a big shift that's happening right now And yet we want to be able and this gives you simultaneously the same kind of a Guaranteed that you had once before but now you get a much more flexible environment So that you can now deliver it in many different ways according to what the consumer wants to achieve Classic example is home DVRs sort of like this because I think we can all relate to this We probably all have had DVRs in our home with a with a clock flashing on and off because we never could set it or whatever And we had to record our video and we had to hope that we received a disc fill up And this would seems fine because it replaced our VCRs Which is what we had in the past but now we're seeing these things moving to the cloud And there's a lot of very good reasons for why you want it up in the cloud one you're gonna have an infinitely large disc You don't have to worry about your disc filling up From the service provider point of view they don't have to keep all the separate videos for every single person there They can actually find ways to be much more economical and just mark the fact that you bought that and not even be able to Essentially store it for you individually Given certain digital rights Concerns around that and they also then can be able to deliver it to you whether you're in your home or not So this cloud-based delivery I think it's something that we're seeing come on very strong because it provides customers advantage and Advantages for the service providers as well. So this obviously now becomes a software model instead of the hardware model of the home DVR And this is the kind of thing that we're seeing so we want to be able to to have these kinds of new services being able to Delivered in this kind of a framework So when we do this we start to see again some of the traditional functions That we would have in a video kind of pipeline turning into software And here's where we're talking about it. It's kind of service containers that can be chained together I just press something There we go And what you have to do now so you're gonna start hearing a lot of people talking about orchestration Orchestration is how do you bring up these services? How do you scale individual services? What if you have all of a sudden the peak demand because of some event and you need to do one of these services expand that? To have a much more capacity For the increase in demand and you need to be able to have workflow managers that can that can move this through the system here so in Concert with that then we're seeing something called docker Kubernetes Mesa coro s this is the containerization of the world that we're seeing because developers now are finding Emulating a virtual computer by having a virtual machine is a very heavyweight thing to to carry around and it doesn't Provide much advantage to the individual developer It's nice because you can actually run a Windows machine on top of a Mac or anything else because we virtualize the computer But when we actually have a simpler environment We want to be able to talk about how do we package an application together and in the packaging of that application be able to run that as a Container anywhere you want to it's something that's tremendously advantage For the new application developer So I think we're seeing these new models now of developers of how we develop applications come about To enable these new kinds of services and increase again the speed of service delivery And for that reason I was to then turn it over at this time to Ken Owens to talk about what Cisco is doing also in terms of now containers There's many people and I think that Jonathan even in keynote Mentioned that some people think of cloud computing and containers as being a kind of war with each other or that you have to choose between virtual machines and containers My view is it's a lot more nuanced than that in fact that we're going to see all these things together What we're talking about particularly with open stack is is the orchestration in the management of virtualized resources and whether that be a container running on a virtual machine or a container Running on bare metal because an open stack we can also orchestrate bare metal It all becomes a single kind of plane that we can talk about as this new kind of cloud platform And so the age of like sort debating whether it's a container or virtual machine really becomes one of a developer choice Whatever they would like to be able to developing that's what we need to be able to support through that I'd like to turn it over to Ken. Thanks Lou So when you kind of think about what Lou was just describing why do you need a new model, right? I think it's not really so much a new model It's just kind of making open stack the platform that open stack was designed to be and so I kind of look at this from two Different areas right there's disruption happening in the enterprise where you're looking at how do you move from kind of agile? How do you move to agile development from your traditional waterfall type of development? How do you start developing these cloud components as services as Lou was referring to? It's kind of a natural evolution. I think from a developer standpoint to look at Here's how I like to develop my services How do I deploy these as services and so it's not much of a change from a developer standpoint from an infrastructure standpoint you want to offer services and so Open stack is a nice platform that allows you to create services and expose those services through API's and so it's a good fit between the two There's also this model of you know, like Lou was referring to how do you determine if you want virtual machines or bare metal or some other In between solution right and I think Most developers don't care right? They just want to get their application out as fast as they can they want to get it some feedback and they want To be able to modify and update their application based on the feedback they get from the market and so Kind of doing this hybrid dev ops where you can develop your application the way you develop it today Deploy it into an open-stat cloud see how the performance of that application is tweak the application performance to improve the performance Modify the underlying infrastructure if you need to to provide better performance It's kind of what I call this hybrid dev ops model the other thing that was seen is Digital disruption is definitely impacting a lot of enterprises today And so how do you kind of look at your your infrastructure as a platform instead of as a set of? Compute and network and storage and security constraints right? How do you kind of look at the commodization of infrastructure? Which is probably not the best thing for your business in the end if you're on low quality low-grade hardware How do you really differentiate your service in the marketplace if you you're building on top of a bunch of you know basic commodity hardware The other thing is around security and privacy so Containerization in and of itself is not secure But when you look at adding security and network capabilities to containers You can have a much more secure and a much lower fault domain when you deploy things as services So just you know a couple quick size into the what developers in the enterprise are telling us the sort of two camps You have your cloud native guys that are Either developing today for the cloud or for mobile or you or the traditional app developers moving to cloud native And you have your data scientists that are looking at how do they take the analytics that are going to help them make real-time business decisions Based on the application So the analytics guys are concerned more about structured data versus unstructured data and how you collect that Kind more like a Kubernetes model The cloud native guys are more interested in continuous delivery and mobile first development more the mazo scheduling model And then enterprise IT is sort of stuck in the middle of the two and and they don't really understand either one that well yet They will but right now. They're kind of like scheduling. What's that? You know, I don't need to schedule I'm more worried about how I provide security and governance and how do I connect you into my support systems? And there's a time figure out how to support DevOps or work in a DevOps model and so When you think about these three different, you know user groups I can't think of them in terms of personas, right? And so you have your developers at sort of the number one person I think about the most and they're really interested in how do I get this application from? Concept to production as quick as I can and not underline understanding under anything about the underlying infrastructure is their mindset You have the IT admins who want to kind of create environments that can be consumed as a service And they're trying to create this environment that the dev developers are going to leverage to take their application from concept to production And then there's a few other, you know personas you can think about like the CIO and the business owner But if you focus on the developer and the admin to use cases you hit most of the use cases that CIO and the business owner care about and then what do they want to do? They want to basically build their application deploy that application and then run it and so when you think about the build piece I think most of The philosophy in the in the developer environment today is don't make me or ask me to change how I like to develop So I have a framework I develop and I have platforms I like to use just let me continue to develop the way I like to develop From a deploy standpoint right now if you think about the biggest issue I think facing, you know open stack and other clouds today is that you have to write to those specific APIs and As Lou mentioned in his earlier conversation The latest version of code is the most stable and so if you are writing the open stack You have to write to the latest version of open stack every six months. You're changing your code or Potentially changing your code. No, I'd always change it But more than likely you're gonna have to change your code at least validate your code again And so when you think about how do you kind of develop your code once and Then just kind of deploy it in a CICD model afterwards. That's sort of the layer That we can write above open stack earlier Lou caught at the sandwich, right? So we're kind of looking at above the top part of the sandwich. How do we kind of create a platform? Interface that allows you to kind of write your code to interface with not just open stack but other cloud technologies And that's something we call mantle Mantel IO is the website and then from a run standpoint Most developers want to understand how their applications and those services they depend on are performing and behaving and so Creating a run environment that lets you from a single interface look at how your applications performing across multiple clouds across multiple services is a key part of the framework and then tying in some of the application intent policy stuff from APIC and from other Controllers in the environment and Tosca being able to give you a single interface that lets you look at your application From one console versus multiple consoles is a key part of the interface And that's what we call project shipped and so project shipped is a curated experience It's available at shipped IO Cisco shipped IO It's tied in with these components you see today. So we're trying to hit the Kind of the beginning part of the development cycle letting developers develop the way they like to develop with GitHub integration We have build packs for the most popular code In use today also Eclipse We're using vagrant to kind of run in a local IDE on your laptop, which is what most developers do anyway We're adding OpenShift and Cloud Foundry support in the next release of project shipped We're using drone for CICD today as well as Terraform to kind of give you an abstraction layer That lets you run across multiple clouds. And so today we support Amazon and Google Digital Ocean Cisco's cloud Cisco's MetaCloud and Along with Rackspace and VMware cloud so completely abstracting the underlying Infrastructure so you don't have to worry about writing to specific infrastructure APIs Although we don't completely hide them and so if you wanted to write, you know to an APIC controller We'd let you write directly to an APIC controller as well So we will kind of give you the a choice of taking the abstraction layer and not having to understand the underlying infrastructure Or if you want to really write to the underlying infrastructure We'll give you access to the underlying infrastructure APIs as well From the orchestration and scheduling standpoint. We're supporting both Mezos Marathon zookeeper and Kubernetes. So we're giving you a choice of all of those. We use console to give you name names namespace and tie it into DNS so each container gets an IP address You can route between containers in your services So it's a making network and security like a key aspect of every container And then from the service assurance side we're adding a lot of data analytics pieces that are being asked for by the data scientists community So things like elk the elk sack isn't is included log stash and zoom data for kind of Analyzing and understanding the different log inputs of your application and then like elastic so should give you You know this sociability of that of that data so We think about what mantle really is it's trying to give you that glue that every project has to kind of use today to Kind of build their application It's an end-to-end batteries included solution It's designed from from day one to kind of give you all the different components of the open source community in an integrated manner It's an extensive also as new components come out. We'll be able to add those components in very easily. It's an open source initiative within Cisco Well, it's on GitHub at Microservices infrastructure and Matto IO if you want to get to it directly from the website And what we're really doing is just basically taking what every Enterprise developers doing today on their own per team per project And so instead of having to kind of build a platform from scratch every time We're kind of giving you a basic infrastructure or microservices infrastructure as a service if you will That can run inside the data center and it can be deployed across any external cloud The architecture is pretty simple. It's Basically, we have control nodes and resource nodes the control nodes give you the basic, you know Scheduling services around How you set DNS up? How do you manage your resources? How do you manage your namespace the resource nodes that you do the on? the local destination scheduling We run it in a single data center environment and we can run across multiple data centers Which is sort of the the key thing that we focused on with Mantle and Today, I had to we support multiple clouds The the key thing I want to get to is the networking piece It was mentioned earlier and so to us networking is not just not just because was Cisco is important But networking is always important for any application, right? And so We're gonna make it a first-class citizen of containers today it runs each container has its own IP address Where we seamlessly integrate with different? networking protocols and to support BGP across the different container spaces We have basically created a unified platform for for networking across Kubernetes and workloads consuming meso services and from a policy standpoint we Releasing in the next few weeks with working with Congress. We're doing here in open stack We have a blueprint available to kind of let you do application policy intent, which? basically takes the whole definition of your application from a Set of like specific request to more of a I would like to have high CPU I need better networking or I need better such security Making it very simple to kind of define what you are looking for what you care about the most and then we can provide the best Solution underneath the covers to support that So actually I we wanted to leave some time for question So if you have any questions, please make your way up to the microphone But but actually can I had I had one question which is that if we look back at for example the shift? Diagram you've got very very large number of open source Systems that are there that must be a huge learning Challenge for many of the developers and they have you in in you're working with different developers developing these microservices How do they keep up with all of that? There's a lot of different things that they have to learn Yeah, they they they don't do a great job of keeping up with all the different open source projects And you think about open stack is a lot of open source projects within itself, right? So when you when you add the infrastructure pieces you try to figure out all the integrations to the back office and to the support systems that an IT enterprise has to support They they are really struggling to do this every single release And so trying to provide a single platform that is above the stack Right and integrating into different stacks that are below that API interface They they are very excited about being able to just leverage something that's open We're not locking them into any one technology We're not forcing them down a path first decision that they're making yeah They're gonna be working above open stack at the application layer Where they actually are creating the new microservices and everything right and therefore there's a whole lot of tooling that they have to learn There which looks very familiar to the tooling that open stack developers See it's all right hub and Garrett and everything else like that I think but it's done at that application layer and they don't have to worry about what what's actually running the platform itself Yeah, I think that it whether it's whether it's open stack on bare metal open stack on you know a KVM or it's You know AMIs it to develop but they really don't want to think about that They just want to write the code publish it Get this feedback from the community as fast as they can on how well their application meets the business needs Right and then make changes from that because again, they're they're in the business to make money for their company I'm in the business to make open stack better right right right right right in fact We should make just mention of two other open stack projects related to containers, which is Magnum and Cola And how does how does what we do with mantle work with with Magnum? For example, which is the the way that we are deploying containers So we've been working really closely with Magnum and we have two contributors on the Magnum project It's our work is tied in very closely, but we're again. What kind of? That's one implementation that we're trying to support but there's others that we also support okay It just just in terms of the audience. There's a Magnum is a project And a part of a big tent For open stack whereby we can you can bring up Kubernetes you can bring up me So and it's designed to be able to deliver it now in this latest release or whatever through even horizon You can start to bring up now this cluster that you need two of these containers And so immediately the the developer is only interested in working at the container level can do that And that can be run either on virtual machines or on bare metal nodes through ironic as well So we're seeing now again that usage of containers for microservices sitting on top of open stack Which can then support both virtual machines and containers because in a particularly I think then the the individual customer gets to make the choice there if they are actually Want isolation between all of the tenants? They perhaps need to run containers on top of virtual machines where we have the isolation the multi-tenancy provided by By the by the hypervisors Whereby if they are essentially running as a single tenant with lots of projects lots of applications lots of services They can start to think about using bare metal using ironic to run Magnum as well exactly nice Any other questions from the audience Gary? How are we doing on time? Okay, well, and I will be up here if you want to ask any further questions, but if not, thank you very much Thank you and can and We invite you to come down to the Cisco booth in the marketplace We're in space P3 come on down continue the conversation asked questions will be there As the comedians would say we'll be there all week. So come on down. We'll see you there. Thank you very much everybody. Thank you