 All right. Welcome everybody. My name is Neela Jacques. I'm the executive director of the Open Daylight Project and The idea of this panel came up because what we what I've been seeing is that while at this point You would think that everyone in the world understands SDN and FV the differences where we are Or at least it seems that way if you if you religiously go in and read every possible thing that you can on the topics every day But I was forced to realize that not everybody necessarily does and so I've been fielding a lot of questions around Where are we on SDN? Where are we on NFV? How are people using it? How real is it? When is it going to be real? Is it is it carrier grade? Does it matter to enterprises? There are all these questions that I find myself fielding over and over again And so rather than get up here and tell you my thought About all of those questions. I thought a much better way would be to put together a panel with a range of individuals who represent different parts of the industry who can give you a feet on the ground view of Where we are in terms of SDN and NFV? There's been a couple changes to the panel that I want to highlight Chris Luke from Comcast ended up not being able to make it at the last minute and Alex Zhang from China Mobile is here to Replace him. We also had another last-minute cancellation Margaret Kiyosi had a medical issue and had to leave urgently and so we have Brian Sullivan here from AT&T replacing So the first thing that I wanted to do is give you a chance to get to know each of these people They're all pretty impressive in their own right some of them In fact wear a couple different hats and I'll ask them to share sort of the different hats that they wear Brian Can you go first? Thank you, Nila. So I work inside the AT&T cloud architecture team Toby is actually the person I work for and we're involved in sort of coordinating architecting motivating the the solutions that we ultimately implement inside the AT&T integrated cloud and domain 2 and My particular role is helping to facilitate open-source engagement inside the open platform for NFV and you know open daylight and open stack etc and Basically helping thread the the AT the broader AT&T team. That's that's doing developer type things into it. You know a strategy Excellent. Thank you Chris Price My name is Chris Price. I work in Ericsson CTO office focused on open-source technologies for SD and cloud and NFV That's my that's my salaried cap. I also I also work within the OP NFV Project, I chair the technical community there and I also attend the technical community Committee of the open daylight project, which is chaired by the gentleman next to me. Thank you Colin Dixon So as I said as Nila said, I'm Colin Dixon I work for Brocade, but I spend almost all of my time working upstream on a combination of Open daylight and a bunch of other things I chair the technical steering committee, which means I'm responsible for things going wrong and have no ability to control it So that's a lot of fun if you want the job. Let me know but I also spend a bunch of my time talking with customers and helping to architect and Support Brocade's distributions not only of open daylight But also of tacker and a couple of other open-source projects and I sit inside of sort of our open-source product team at brocade Good morning, everybody. My name is Alex. I'm from China mobile for me, I actually are working in the Silicon Valley office We have the advanced the research lab there. I'm working on the SDN and the NFV technologies I'm also part of China mobile's SDM and the NFV program called NovoNet So this is China mobile's program to build our next generation network We're gonna use a lot of SDN and NFV technologies and we also embracing the open-source projects So that's why we are here for this summit For myself, I'm pretty active in the open-source projects I'm the one of the members in the advisory group of open daylight and also the vice chair of the Operator area for ONF and I'm also active in the open FV and so Very good to be here perfect Thanks, so you see here on this panel what we have is two people who work for user Organizations to individuals who happen to work for vendors although they work very closely within the upstream Community and I think this actually represents a lot of what we're seeing in open stack in open daylight and no pnfv I think where we are today is we're in a place where the communities have really come together and then users sit side-by-side With developers whether they work for vendors or whether they work for research institutions And so with that I wanted to start with probably the most burning question that I have been getting recently Which is lots of people I think understand technically what SDN and NFV are and yes, they are different But one of the questions I keep getting asked is well, what is it good for? What are people actually doing with SDN and FV? Is it just a set of pox where people are doing are indeed at trying so to see if this technology might be good Might be useful for something or people solving real problems I wanted to start with Colin as you manage the overall developer community in open-day life I'm sure one of the questions that often comes up is what problem are we solving so tell us What are you seeing in terms of the primary use cases out there for SDN and NFV right now? So that's one of the questions we've tried to answer since inception and we tried to answer it The first time in hydrogen and we got it really really wrong We released a service provider addition, you know a data center addition and something else and Nobody found any of them useful Since then we've had a user advisory group that's been able to give us a ton of feedback We've also seen more people pick it up and run with it and the use cases are incredibly broad which Speaks really well the fact I mean open daylight has a just a huge number of different southbound drivers and Applications and models inside of it so you can it's really a Swiss army knife to do whatever it is that you want to do But you know specific use cases we've seen we've seen a lot of data center interconnect, which is more SDN than FV But you know certainly data center connects when you start doing really sort of SD when and you end up with optimizer sitting at the end Starts to have to do life cycle management of VNF in order to try and do compression You know mitigation of loss collecting alternate paths you do data center connect So that tends a bleed between the two we've seen network virtualization be a significant use case Surprisingly a lot of network virtualization is in support of not necessarily connecting the end points, but in Support of NFV which is basically how do you you how do you create the virtual network that is your service chain? Through the VNF that you provisioned even if the end points aren't under say open stacks control So how do you look at open stack is basically one giant middle box which you can then use to control everything You can then orchestrate paths through even though the end points aren't resident within an open stack controlled environment And we've seen pretty much everything else people trying to replace their wands with software to find ones in order to do you know traffic engineering internally even without like when they have direct control over the links things kind of like rules before and All the way to straight up just you know life cycle management of your VNF's Which is actually still really really really hard the tacker project is taking stabs at it But how you actually go from you know spinning it up We're you know updating it getting its configuration right winding it down making sure you can scale it out Life cycle management of VNF's is really really tough And I think open daylight provides a reasonable layer on top of standard VNFM It's just sort of talk about that and how to make that migration simple. That's a bunch of the use cases I mean there's there's many more than that But hopefully gives you some sampling of what I see people using it for in practice great It's a column I think what I got out of that is you're seeing a broad range of things This is technology that can be used to solve many many problems Maybe I'll go to to an end user Brian as he's picking up his his glass, you know at AT&T AT&T Obviously has a network that is quite important to it And there are many places you could start with SDN and NFV Can you talk a little bit about what are some of the places that you've started and why have you started there? Okay Well our first major product based on the SDN concept is network on demand And we started there because basically it's it's fundamental to our very large Distributed network and it's a key way in which enterprise customers. You'll get service from us You know we we are you know working on NFV and preparing VNF's to be rolled out, you know And it's basically working on the cloud architecture and cloud elements of the AIC But but to start with what we fundamentally know as networking, right? We started with with network on demand and it's it's actually working out quite well for us Margaret gave me a statistic. We have 450 customers On this over 450 customers and 2,000 ports on our network on demand service it enables us to you know begin with provisioning circuits and you know in in near real-time and drastically, you know reduce the amount of time it takes to to to turn things up and and actually show the promise of an automated control plane for networks So that's probably the number one and following after that, you know, there'll be quite a lot of virtualization of Existing functions in insider data centers, etc. That will be rolling out. Great. Thank you Chris and Alex is there anything you want to add from your perspectives? Yeah, okay, so I'll add a couple for use cases The one thing is similar to Brian said that in China mobile like I mentioned earlier We do have a program called a Norvernet. All right, so we have Designed quite a few what would be the very important use cases to you know deploy SDN Technology and so one of them is what we call a transport network So this is where we want to Provide this network on demand and they establish the Dynamically connection between two customer sites to right here. So this is one of the kind of a similar scenario So we do look at it the Open daylight and as the base controller so we can build application on top of that So this is one use case the other use case actually is very important for us is Some are similar to what I just mentioned the data center So we are evolving our data center. We do have a public You know public, you know cloud services and the virtual private cloud services So we are offering in mainland China. So we like to enhance Improve the networking part of this the cloud service or the infrastructure. So we use the SDN controller to develop a new technology I based on the the virtual Underly and in the overlay and to control it more efficiently and to allow us to have more elastic networking Capabilities so that we can evolve our data center to the next generation Great. Well question for Chris price now We're here at the open stack summit, right and we've been talking a lot about SDN controllers and we'll talk more about SDN controllers But let's talk a little bit about open stack You're the chair of the TSC for the opnfe project and I know that there's been Tremendous pressure I think to test how can people leverage open stack specifically as well as some of the components under an open stack Infrastructure to deliver virtual network virtualized network functions Can you talk a little bit about the role of open stack specifically within NFV? Sure, I mean so so open stack is More or less the preeminent infrastructure management suite that we have in the industry from an open source reference perspective in opnfv we we have an Essentially a vision to come out and create an executable platform a reference platform that provides a hosting environment for network functions which Can be extended into the network and can extend to an edge environment Open stack is the place that we come to it's the place with the broadest community It's the place with with the most I guess traction and and velocity when it comes to future development To help us address those use cases. So when we want to deploy a network function of VNF We can do so knowing that Things like nova and over scheduling new mappinning types of capabilities that we need in order to get the best Performance out of the functions that we're trying to deploy are available to us We know that we can do networking. We know we can use neutron if we want to layer to network We know we can use open daylight if we want to start to do LSO API for connectivity services to an edge For instance, we know that we can we can extend and leverage these types of capabilities and from a Coordination or a management and orchestration perspective we can leverage interfaces from both Open stack and for instance open daylights northbound interface to coalesce this into a set of commands that allow us to essentially From an NFV perspective get it get a software interface towards both the network and the cloud environment All right follow-up question on that Open stack wasn't necessarily designed for an FV to begin with But it can certainly be very useful to an FV are the needs the same are they exactly the same? Are there some additional needs can you compare and contrast a little bit or any others who have insight into this? Around the needs on open stack from an NFV use case versus a typical cloud IAS use case Yeah, so I mean I think to begin with we started with this telco cloud is 5 9's telco cloud is better Performance telco cloud is all the things that an enterprise cloud wants to be anyway I think we were mistaken when we started to articulate that as what is a telco cloud I think from an operator's perspective and these guys can correct me if I go through rails But but my understanding from an operator's perspective is it's not about having a cloud which provides you a virtual Environment to run a function. It's about virtualizing your end to an infrastructure about creating an environment whereby I can Instanciate a service across my network which which may end up with functions in a car running on a freeway that I'm deploying and having life Cycle management of in real time Through my virtualized infrastructure. I mean these types of use cases. This is what NFV is all about. It's not about building a data center It's about Creating the capability for that data center to extend out into the network and to extend in a normalized and standard way I mean Google Google's great Google has a sandbox the world's biggest sandbox and they can do what they like If you want China mobile and AT&T and Verizon and everyone else to be able to do those types of things We need normalized standard systems and we need to start to deal with those same types of use cases We need to create an environment Where I can treat my entire network as a virtual platform. I think that's NFV right anyone else want to add any color to that Yeah, I think to me like we of course we we Embracing the open source projects. I believe open stack, you know position very well in the resource management, right if we have a telco worker loads we want to Support it by the network function virtualization or vnfs. Yes, only I will go with the open stack to begin with All right, so this allow us to control the resource So this is what I call it a computing part, but there is also the connecting part, right? Which is so we used to call the networking so so I believe we recently are Looking at the SDN controllers to provide us more advanced the capabilities to allow us to do the networking part So I think the open stack Networking part we have a new tone, you know API, you know ML 2 is very popular, you know why they use but on the other hand they want to look at the end-to-end Chris mentioned and then the networking needs or even the Oxygen you know needs and then we have we need to have a more advanced in networking capabilities This is where we believe either we can have you know these capabilities from SDN controller or controllers Such as ODF or we need to largely enhance the networking module of the open stack Right, so this I believe there are some projects going on right to add the L3 VPN capabilities to the open stack So this is actually pretty important, right? So you can push the scalable layer three solutions All the way to the end so these are the kind of things that are at least though we are looking at it So Alex what you're saying is that you're investing in SDN and NFV across your entire network and so you don't want your Open stack based infrastructure as a service cloud to be a separate silo. Yeah, that's yeah, I certainly this is the key point one of the key points We do want to have you know the one integrated system that will allow us to do everything and not only the networking part Right, and so it's not gonna work. You know, we have individual components sitting in the silo, right? In fact, you know, we actually turn on mobile We are pushing you know open source in orchestration system and you know effortful for bring all things together, right? So there will be a session tomorrow morning and you know nine o'clock So so please join that session of things in the next next room, right? So we're gonna talk about open source in orchestration. Yeah So so the key point is still the same that we do want to have for you know All these the you know component and to work or system to work together, right? So good I want to actually want to double-click on something. I'll go to Colin next and then Ryan a question I often get from the open stack community is You know open stack is complex enough you're talking about bringing in an SDN controller doesn't that just add complexity? Why do we need it? Why can't we just use ML to I'm sure you get that question a lot too Colin? Maybe you can take the first crack at it from a developer perspective And then we can get Brian to take the user perspective on that one So sure so ML 2 is and you know neutron in general the API's are are are great for The reference implications of them are really fantastic for Modestly-sized deployments which to be honest are most of them So if you have like a single rack of servers or maybe two and you have two tours The truth of the matter is Nova Network probably worked fine You know, I mean that's just I'm just being honest And so like, you know, if you needed separation you could resume the lands as the network gets more complex You start to span across Geographic areas you start to have actual bottlenecks and constraints you start to have to thinking about routing through service functions And so pathing becomes complicated and non-trivial You don't know whether you are leaving a vnf or entering a vnf necessarily when you sort of come on to a v-switch And so starting to track all of that stuff tends to strain The interfaces and implementations that we have today And for a better or for worse the most robust implementations we have how to deal with that Sit outside of open stack and SDN controllers like you know open daylight onos and the like And so if you're curious, so that's sort of when I think you start needing to talk about Using an SDN controller and I don't really think I mean the difference between open daylight and being a second Project that sits over to the side and being a project with an open stack As a developer, I don't find it compellingly different. They're slightly different communities, but there's a lot of overlap But the level of complexity that you add on I don't think is wildly different And so it's really when the network starts to be an element You have to think about and reason about you have to think about it reason about it And you might as well use the tools which have evolved in order to do that As opposed to sort of pretending it doesn't exist Which works really well at small scales and not so well at large scales Great Brian, can you give us the user perspective on that question? Yeah, well there was a good bit of user perspective in what Colin said as well But but I would put it this way. I mean, you know if someone said, you know Why do you need this thing beyond what open stack can do? You know, I would say okay first of all we've got Dozens of data centers. We've got thousands of central offices, right? And we're gonna put you know clouds in all these places, right? And if if open stack was just you know Shipping bits out of the workloads, you know into a port somewhere on some router inside the edge of that, you know Point of presence Basically the network as we know it today would be exactly the same it would be static it would be provisioned It couldn't react dynamically, right? So that's the fundamental thing is that the this this very broad dynamic Network environment requires something which is designed specifically to address those types of issues that that's the fundamental thing Why we need a controller? The second thing is is that network functions as a you know, they they call them network functions for a reason, right? They really have a life cycle they they need be managed, right and a controller provides us a way specifically to design You know using a for example in open daylight a modeling You know language to design the life cycle management of that network function in ways that open stack doesn't enable us to do And in the end, you know, we need both functions We need cloud and we need network working together through you know an orchestrator To make things happen in the right time, etc And we we need open stack to be really really good at what it does for example in the last question My response would have been NFV on open stack requires it to be just massively scalable Multi-site, etc There are a number of things we need in open stack to do really well for compute and storage and what it does best Right and the the SDN things we need to focus upon you know that that whole problem domain You're really separately and a controller does that for us, right? So I think one of the things that I'm hearing from a number of you has to do with scale Right and is that whether we're talking about SDN or whether we're talking about an FV when you hit levels of scale That are high enough that one or two people can't do it themselves You need some degree of automation to be able to get the agility that you want and certainly in carrier cloud You have that scale. I was recently in China at Baidu and I was knocked flat by the scale That someone like Baidu has to have And therefore this is why they're turning to solutions like or technologies like open daylight to be able to To manage what might end up being 90,000 top-of-rack switches in their environment And I just want to be really clear that like open daylight is it we do We are goal in life is to be able to be the best networking provider for open stack in Existence and so if you have ways that we could do that better that would be fantastic But we're not out here trying to create new APIs and saying that neutron is the wrong approach We have a complete implementation of the neutron APIs. We're working with the networking SFC guys inside of open stack So really this is not a like you know look we understand how to write these APIs better than everybody else So just come listen to us This is really just you know us saying we have a set of primitives underneath that we can use to help implement the APIs Which people find useful in the open stack community in the NFV community? And so think of us as you know a sister project open stack that happens to not sit an open stack org But really we're trying in every conceivable way that we can to be as close to it I mean just supporting those APIs when we need additional features We develop them on the side and we try and figure out the right way to integrate them upstream And that's really what Chris and opnfv Are doing really excellent jobs of which is trying to be that broker So at last you get the idea that we're saying you know just don't worry about networking We've got that walk away from it. That's it couldn't be further from the truth a Question for Chris price now One of the questions that I have heard Asked over and over again has to do with maturity right so one of the questions You're looking at implementing open stack and open daylight or another controller together Especially in an NFV environment is these networks matter right and you may only get one shot maybe two But the cost of failure is really really high So people have been asking this question of you know when is a time how ready are we where where do we start? And so I'm curious as as the chair of the technical steering committee for opnfv your view of where we are we started to see Red Hat we started to see canonical Ubuntu package a controller open daylight with with open stack today you now an opnfv Have gone through two releases I believe can you talk a little bit about where you think we are in terms of the maturity curve of a solution that includes those two components So in opnv we we package a number of SDN controllers within within the cloud environment We take neutron we can provide a cloud with neutron networking. We provide an open daylight Based networking and we include own also an open contrail and then and we can include any Networking solution that we like what we do is we we put it towards paces we run For any given release we're running millions and millions of test cases. We're pulling upstream test cases running our running our own test cases We try and do things at scale well We try and do things in ha and high availability Deployments on bare metal with top of rack switches so that we're not just running things in a cloud and Expecting things to work We run it in the situation where it has to talk to an actual switch and where we're going to run into any Erase conditions because we're running a three-blade high availability cluster We do that repetitively we did that in in our first release Arno and things didn't go so great The maturity of the solutions weren't ideal and we learned a lot We came back to the open daylight community We came back to the open stack community and we talked through what were problems We were finding and we did a stable release of Arno which improved things considerably you could actually use the solution And then we moved to Brahmaputra where where we actually went through the next big iteration We pulled in Liberty. We pulled in beryllium for open daylight specifically and With the feedback that we'd given and and with the test cases and verifications that we'd we'd actually worked with We now see that we can take these technologies. We can deploy You know I miss every single time we deploy an open stack Instance in opnfv and we've deployed over 2,000 of them so far this year Every single time we do that we spin up a virtual IMS solution on that just to make sure that it works And that's just a baseline for us if it can't carry a Telco vnf then it's not passing our test cases and an open daylight to be honest Go back a year it couldn't it really couldn't pull up an IMS system It would fail now it can now now we see that we have the maturity in in the integration maturity in the software that we can Start to exercise and leverage the types of workloads that we want to actually support Anybody want to add any color on that? I've got another question teed up then The last sort of one that I want to ask and then I'm going to ask folks to get up and and give us some questions some User question or sorry some audience questions I'm often asked this question around the role of users in an open source project I think those of you who've been around to a few open stack summits in the early days There was a question of where are the users this seems like a developer or a vendor led Project and I think open stack has gone a long way with the super user program and really bringing end users into the fold this same criticism was thrown out at open daylight and an opnf e in the early days and I'm curious for From each of your perspective of where you are can you talk about the end user? Developer community that has been created. Do you see these communities as? Truly working together the user community and the developer vendor community And what do you see as the role of an end user in an open source project like open daylight opnfv or open stack? So yeah, I think the The fundamental reason why opnfv was formed was that as users as people working inside Standards on NFV right in Etsy we saw it wasn't happening fast enough, right? And so somebody who's just working on technology developing something They probably you know they like it and it's interested in and they do good things But but nobody's sitting there stamping their foot going can we have it yet, right? And so that was a very end user focused thing Let's get it done and and and then in the process we brought together, you know the the implementers the the you know the large operators Etc as a as a community of developers and users together and this week in In open stack for example the in the ops summit that happened on Monday It was very clear that you know we we share a lot as an end user group with the very broad set of operators of open stack clouds In fact many of these operators are out there, you know Reporting on things that they find their experiences, etc And not necessarily having like like some of the the big telcos or some of the you know the big operators not necessarily having the the the time or the resources actually go fix them so as an end user community those of us in opnv and and those large operators inside Open stack we really need to represent this broader end user community And that's I think that that kind of sensibility is forming and and we're going to be Actually taking a lot of action to to you know to drive solutions per those those concerns in future So I think that's the role of an end user group is to coalesce around the real priorities And then work and find the people who could just get it done great I'll chime in on the opnv side I think as Brian alluded to when we formed the opnv we formed it with it with the user community in mind And an active user community and Brian himself is a is a ptl for two projects one on one on api normalization and one on policy and automation And and he just comes and he gets stuff done and and all of a sudden we have a platform which is supporting policy and and It's just a matter of getting in there and getting things done. We have we have entity document We have orange. We have we have China mobile all putting resources in all leading projects all all establishing Trajectories for technology development within our community alongside vendor partners I think that's it's one of the big success stories of opnv and I'm gonna stop talking I'm actually gonna pause because we have only a few minutes and I'd love to get some questions I see we've got one person in the mic. So sorry about that guys Hi Nela Scott Fulton with the new stack in the Containerization space there are a number of ways in which a number of the orchestrators will utilize SDN to scale workloads out and I'm wondering if if you gentlemen have any opinions as to how Open stack uses NFV to enable end-to-end Orchestration and end-to-end resource orchestration in a way that is superior to the ways that The containerization space. I'm thinking of a mess of sphere DC us and Kubernetes are using and Or is there a way that some of these orchestration tools in the container space containerization space are Teaching us new things that we can incorporate into open stack and NFV All right. Thank you for the question and if you guys want to take a crack at this one So so so as I understand the question was essentially we have these container orchestration tools which Look like open stack if you squint Kubernetes mazes fear Kings like that But they are different in a variety of ways and sort of what can we learn from that? What could an SDN and NFV bring to it? So I think the thing which they've nailed in a way that has frustrated the R side has been the providing a unified Abstraction to the whole system so the idea that the data center as a computer as opposed to the data center as a way to write A for loop around 75,000 elements They got better than we have. I think that's something we should take away. We should learn and we should move forward with That being said I think their software only view of the world Or you know container only view of the world tends to limit some of the pieces So the idea that you might want to be able to have an NFV orchestration layer Which could declare load balancers in between your different services, which might be p&f's and be physical load balancers They might be v&f's there'd be a combination of the two bursting from physical to virtual and actually treating the network as an element that is Real and perhaps complicated and difficult to schedule is something that they haven't done very well And something where it is that we could help and we certainly need to figure out how to better serve them So I don't know if that answers the question, but we have another question over here You're speaking my language. I work for a telco carrier. We're just now starting our virtuals a virtualized services from kind of the ground up and You know, we're getting pushed by say accountants to virtualize everything and they just walk in and say, okay virtualize and People come to me and they asked me, you know, well, why isn't just going to work tomorrow? And I said because you know open stack as it exists is meant for containerization That's meant for data services. It's meant for this and that but it's not specifically meant for networking so We've come so far in such a short time I mean, how long do you think it's going to take for us to get up to literally a telco grade? SDN fully realized SDN I might actually ask Brian to start on the answer there and then I'll go to you and Brian out just a quick Upfront so John Donovan stands up two years in a row at ONS and says we're gonna hit 75% We'll do it tomorrow right more or less something like that Last week so really very bold statement Obviously, there's a lot of work that has to be done I get a peek under the scenes. It seems like there's hundreds of engineers working on this Can you maybe give a sense for those people who maybe don't have the full resource of AT&T? What is it take to get there and what and where are we from a technology perspective? Well, okay, so What John does is is very essential basically he's he's giving the message to the the market and to us Right that this is a train get on it, right? And if you know if you're not ready to get on at the train will pass you by and hook you and And drag you with it, right? So we are working extremely hard We have a huge number of vnfs that are in the queue to get rolled out into our domain to environment You know that work on demand though. I think showed that SDN is real. It's it's ready for prime time and you know Actually redeploying the the the custom network functions as cloud applications under open stack is is is very very close You know, I can't say specifically what we've done so far But I can tell you is a very fast moving train and it's it's going to reach this destination very soon So I believe the numbers that he closed and I believe Margaret has shared that you were just shy of six percent thus far So 75% was the goal 6% is what they've gone with a lot of people and a lot of work And so I think actually that speaks to one of the pieces which is yes, this is happening Yes, this is real, but it does require as Brian was saying People at the top making it a priority and a lot of people making it a focus of the work that they do Chris, I mean I'll just add when we're not We're not baking the whole pie here. We're building a layer cake. You have open stack today You have the ability to create virtualization Solutions within your network you have the ability to interact with the network today to create interconnects and and actually hit that first target Get that first foundation in place That starts you on this journey then from that you start to build the automation you start to come in with the normalized APIs You start to talk to vendors about how you can how you can improve the way you're operating your network And then you put that second layer on top and and this is it's not a simple journey that the We need that we need that first layer before we even know how we need to build applications in the future I mean a lot of vendors are guessing at this. We think it's gonna look like this. This is the future and Maybe they're right and if they are good luck to them, but I mean it's it's it's a path. It's a it's a journey When when the accountants come in and they say let's get started given the bill of materials for a few hundred data centers and say Okay, let's go I mean this and then then figure out how that looks in your in your network and then figure out how to automate across your network This is this is a journey we have to take So counter to the layer cake thing. I'll just say pick your biggest pain point and try and build the vertical I mean do something do something now if you ask me when is sdn gonna be fully baked and completely realized Probably about 10 years after the next technology is relevant. I mean no, and this is this is how technologies work So pick something real that you can actually do today and do it if you can't demonstrate value quickly You're gonna get lost in the woods Yeah, I'm gonna add something to this 75% go for virtualizing everything and it's good to have a goal. I think 75% Virtualization is a very aggressive goal You know, so China mobile is a big in a carrier, right? We are pretty large scale networking in China So it's not easy to migrate everything to virtualize the world. So it has to be a kind of gradual process And we like to move talking about a speed and calling mention Yeah, we'd like to move would like to you know deploy services as you know quick as possible So so that's why I you know the virtualization of the function It's part of it. But but in this it's gonna be you know Gradual process to get there so in between we may have also the physical functions that we need to you know Kind of plugging and work with right? I think last session the Toby mentioned that the slides and often the X You know function, you know, I physical virtual so So that's kind of part of it. The other important piece is really networking piece I mean, so I mentioned earlier in my comments that you know, I do believe that the SDN controller Whether it's open stack in a new charm, you know, which is served as the networking piece for open stack a lot of you know capabilities for networking is to be to be there and as for example the IP You know networking the VPNs, you know the the overlay tunnels in how do we do this? So how do we you know plug in this? centralize the control and the capability to that and those end-to-end the service You know control right the the quality of service and the access control and even the security issues All needs to be there right in order to provide and then service. Yeah, excellent. Thank you Well, we're out of time. I just want to say a couple closing comments I think you know one of the things that you saw on this panel is four individuals working for four different companies In different roles from each other But who all for the most part tend to come at the same problem from their different perspectives And I think this is one of the greatest parts about open stack is open stack Was one of the first projects since Linux to really bring a broad range of people together Each of which had different interests, but all who thought the collaboration with each other It was a better way rather than being fractured and siloed in the industry we're seeing the same thing happening in open daylight and in OP NFV and The one thing that I know is the problems we're trying to solve Will not be solved with just the people that we have there in the same way as open stack has been growing Substantially and has needed to grow substantially the NFE community in general the SDN community Together with the NFE community also needs to grow so I'm hoping that a few of you having listened To this discussion realizing we don't have all the answers don't walk out going well You know what? I'll come back in two years and see how much progress that they've made We'll come up to one of us and say hey, I'd like to get involved. I have something to add I have something to learn and I have something to add and whether it's because you're a developer joining the development community Whether you're a tester wanting to do it and end user wanting to give your input. We have a great advisory group and open daylight I believe OP NFV is about to create one. So if you're a network architect, there's a great role for you We have many many ways for you to get engaged and be part of making this revolution happen I think at this point we've turned the corner We know it's going to happen But a lot of really good work needs to be done to continue to move us along Towards that so I'd like to invite each of you to just to take a second before you pick up your bag and walk Out and ask yourself am I in a place in which I can and want to add to this effort And if yes come up to one of us give us your card and we'll try and figure out a way to get you involved with that. Thank you