 All right, welcome everyone. That was great. Did you enjoy that? Okay, so we're gonna switch gears a little bit and talk about SDN at 10 and It's it's hard for me to believe personally that SDN is 10 years old We're gonna take a look at the past and the present and where SDN is headed So we have a great panel here and I should be a lot of fun. So Nick I'm gonna start with you and Maybe you could share with the audience. How did SDN get started and where did that term even come from? You know, you got to be careful asking a professor an open-ended question like that because I could talk for hours but You know it when thinking about where SDN came from It's useful to think about what networking was like in 2008 If you wanted to build a network you had to build it from the same equipment that everybody else had You had to use the same equipment that was provided by all the switch and router vendors They were closed proprietary vertically integrated the features were baked in from the beginning everybody had to deal with the same thing Network management was a dirty word If you wanted to manage a network of switches you had to write your own scripts over a lousy cruddy seat CLI and Everybody had their own way of doing it in order to try and make their network different from everybody else's and The equipment vendors had no incentive to help you to do that because they wanted all the intelligence inside the box Because that's how they could extract the most margins from the equipment that they were selling This is not a cynical view. That's actually the right business practice for them to do but for the internet This had a stagnating effect For anyone trying to build a network it made it really hard to do what you wanted So the term the internet has ossified came about largely from guru and from a few others Who were pointing out that the network was unable to change and so as researchers? We really wanted to try and figure and this started as a research project It was how do you introduce change in innovation into the network at Stanford? we started a thing called the clean slate program and That was back in about 2006 2007 and there was a PhD student in that group called Martin casado and Martin had this idea which was hey, why don't we rebuild the Stanford network? Just slightly bold thought Out of switches and access points that we'll build ourselves We're going to replace the 2000 switches that we had in wiring closets every one of them had a CPU on it So there were 200 system administrators 2000 CPUs to control Stanford's network Martin just simply showed that if you lift the control up and out of the switches up Into servers you could replace the 2000 CPUs with one CPU Centrally managed and it would perform exactly how you wanted Be could be administered by about 10 people instead of 200 and you could implement the policies of a large institution Directly in one place centrally administered so we went to talk to a bunch of vendors and said hey This seems like a good idea. We put all of our work into the public domain. Why don't you take this idea around with that? So they got really angry They got redder and redder and redder and we were like this is really strange We seem to have touched a raw nerve here. We got out into the parking lot and said, okay We better double down on this effort So we went back we tripled the size of the research program the clean slate program was started and then in interview a few weeks later Kate Green of MIT technology review coined the term software defined networking And in fact, I have the original paper right here So there is the paper. It was called a thing, right? That's right after by Martin and and yourself and Second paper open flow that was authored exactly 10 years ago April of 2008 and there's the paper that Nick just referred to which is the MIT review paper on software defined networking and There you go. If you didn't know that before that's how the concept of SDN was born so Quiet a journey definitely in terms of how that came about a guru. I want to ask you I remember reading the paper on open flow And it was just mind-blowing when I read it. It was you know, it was new It was disruptive. It stirred a whole debate in the networking industry on Open flow and architecture that you brought about Can you share with the audience? What inspired you to come up with a concept of open flow? And you know, what what was the you know, sort of thinking behind how that came about? So I guess how we came up with the concept I think lion's share of credit really goes to Nick and Martin and Scott But in terms of as Nick mentioned that in 2006 2007 time frame the research community was feeling as if that they the internet was ossified and their ability to Make an impact on internet was very limited They were not succeeding in being able to bring about a new capability the new features and functionality to internet It was becoming harder and harder and actually in the research community We had tried out different ideas in terms of how we can enable and empower the research community to do this And then the simple idea what seemed like a very simple idea to just separate the control plane from the forwarding Define a protocol that is open flow and then be able to Enable research community to build new capabilities and functionality on top of that control plane kind of came about and Caught attention of the research community and made it very very easy for the research community to innovate and interestingly at the time that number of vendors Agreed to put open flow on their boxes open flow agents on their boxes. So very quickly. We were able to get Switched the router that were open flow enabled and then NYSERA and Stanford created this open source controller the first generation NOx and beacon and all of that and they were distributed to the research community and It really caught the imagination of the community And then they were able to create lots of very interesting capabilities on those simple controllers and show that you could kind of bring new Capabilities, there was one more idea that came about right after open flow And that was this idea of flow visor that the idea was that you can slice a production network using open flow and a simple piece of software and then with slices in one slice You could run production network and then another slices you could run Experimental network and show the new capabilities and again in the research community that idea of network slicing and flow visor Got immediate attention and then we were able to deploy open flow enable networks on many campuses around the world And they were able to do experimentation on production network With slicing now it is too bad that that particular concept hasn't really caught on In the industry as much as yet But if you go back and look at the history of open flow and flow visor and all of that all of that happened in something like 2000 I mean late to the you know some 2007 to 2010 2011 time frame and they were at least 15 50 or more institutions that were running open flow with flow visor slicing and Demonstrating production traffic and experimental research traffic running side-by-side on the same infrastructure and how innovations can happen hopefully industry will catch on more and Put some of these ideas into practice. I hope that answers that's great Yeah, so the notion of disaggregating the control plane and the data plane I think brought about a whole new way of networking which is running software on x86 Posts I want to ask Chris this question because I found the intersection of open source and SDN Some what you know what's happening in parallel, but then he started to intersect as networking became open and developers started joining the movement and you know look at how to make these Composable vnfs and sort of decompose the whole notion of a monolithic, you know router and switch so I wanted to ask you how did you see those two movements were to come together and Was that a perfect storm that you know SDN was it was happening in the networking side and while on the cloud side You know there was this whole open-source movement and the developers sort of came together around that well, it starts with some practical issues right engineers you're trying to solve real problems and The the confluence of events. I think was interesting was you on the one hand We had academic community producing a mechanism for creating programmable networks that we hadn't always had so When Nick was describing the CLI managed network infrastructure and how antiquated that is I was thinking that sounds suspiciously like the network infrastructure that we're managing today However building all of this new virtualization Technology and bringing it into into enterprises and and to the to the world at large created a need for a type of network programmability that I think was Happening at the same time as the research and so having tools like Open-source tools like like open v-switch being a great example Having those tools available so that we could build the type of network topologies that we we needed in virtualization Taking advantage of the underlying concepts not flow visor concepts and not doing experimental Networking protocol research, but actually just trying to connect applications to each other and to the rest of the network became a really practical application of of interesting academic research and you know in the beginning So much hype about SDN and disaggregation and and it's all about open-flow and one of the things that I always wanted to say is it's It's not about open-flow. It's not about a protocol. It's about a concept and the concept is about programmability of the network and Open-source is a great way to help develop Skills and advance the industry with with a lot of collaborative effort So taking some core tenants from research Creating open-source projects for people to collaborate around and solve real engineering problems for themselves. I think was That confluence of events and so to me it's a little bit of the virtualization a little bit of academic research coming together Just the right time and then accelerated with with open-source code that we can collaborate on Yeah, and I think it's really helping the networking industry drive a whole new wave of innovation Because I think he opened it up to a huge community of developers with new ideas and And I literally what the cloud anyone can now get some compute resources to start developing You know networking applications, etc. I want to kind of take a look at where SDN is today and There are a lot of networking open-source communities out there now and You know their service providers are you know Essentially reinventing their network and the traditional sort of pop and central office Architecture is completely being transformed. So maybe guru you could Talk a little bit about how that transformation and where it is today I know you're running the court initiative and there's a lot of transformation happening in the Sort of the very, you know Traditional central office space, you know, where's that initiative today? And how do you see that industry transforming? Okay, so Thank you for asking that question as some of you may know card is my favorite topic these days to talk about So I'm happy to talk about that but before I get into that. I also want to acknowledge Especially AT&T and to gentlemen from AT&T one I can see Al Blackburn and Tom ensured who kind of brought that focus on central offices I still remember Al sat me down in one of the conferences and said it is a stupid central office I mean, right? He was telling that focus your attention on central office because that is where Service provider for spent most of their capex and Opex on and that is what needs to be reinvented and he and Tom Ensureds are the people who basically asked us the following question that if we were to do central offices again Clean slate, how would we do that or another way to post that question is if the cloud people were doing central offices? That is Google Amazon or someone how would they build their central offices? And that is what got us down the path of doing cod Now yes, card has gone tremendous traction. There are 20 plus providers around the world that are deploying Cod in the lab settings Some of them want to do field trials and production deployments as well and you may ask why is that the case that all these operators? Kind of cod has captured the imagination of these operators and I think the answers are pretty straightforward We all know that operators are wanting to rebuild the network edge Because 5g is coming Many operators want to do gigabit plus broadband access to their residential customers Okay, and then the edge is where they can innovate new services and revenue generating services To be able to take advantage of 5g take advantage of gigabit plus broadband access and so on and the central offices are very very old And so building this new edge is almost a mandatory as well as what they want to do and Now when they want to think about building this new network edge Obviously, they want to do it with this new software defined networking open source. They want to do it with disaggregation White boxes and all of that that is how they want to do build the new network edge They don't want to do it the old-fashioned way And what called has done is to bring all of this thing together in a single platform or a solution Where you can support gigabit broadband access using merchant silicon white boxes open source You can try to support disaggregated radios with X-RAN and X-RAN controller You can do disaggregation of EPC and virtualize EPC and cod brings all of this thing together in a single platform That one could deploy and that is one of the reasons we are seeing a lot of interest That it is bringing all of these ideas into a single platform and hopefully it will continue Yeah, so I think the the concept of this aggregation is being applied everywhere central office pops data centers And there are multiple abstraction layers now. They're getting formed and you know looking at essentially how to provide a simpler interface To developers. I want to ask you Nick with the p4 initiative that you are driving How is that abstraction layer with the p4 program language really changing the the packet forwarding engine? So I think it's worth to think about the following question And why is it that it takes so long for? networking equipment to add a new feature when VxLan was first defined by VMware and Cisco in 2010 it was four years Before that feature showed up in silicon It's because we've got used to the idea that silicon is fixed function If you can take those things which are fixed in silicon and lift them up and out the protocols and features So that they're defined in software Then software developers will add ideas that we had never thought of about how to simplify Improve make more reliable make more secure the network that they're running today Network behavior the way in which packets flow through a network are determined by chip designers and chip designers don't operate big networks so few of us got together in about 2013 from Google and Microsoft Intel barefoot and a few others to To try and figure out Could you define a language which ended up being called p4 for defining the forwarding behavior of packets in a network? We had this sense that SDN had made the put the control plane in the hands of those who own and operate networks How could you put the forwarding behavior in their hands, too? How could you make it so that somebody could change the behavior of the way that packets were processed? Because if you can't change the way that packets are processed You're not really in control of the network because all the network does is pass packets around really at the end of the day That's all it does it passes packets around and process them a little bit as they go by if you can't change the behavior Then you're not in control So the p4 language was designed to allow that lifting up and out so that other people the software developers Some people refer to it as a as a Cambrian explosion that you would expect to take place That doesn't mean that everything gets far more complicated in many cases people will use it to reduce the complexity down to what they actually need So, how do you see it switch gears and move to the future? How do you see you know with these different initiatives that are happening? You know cord and p4 programmable language providing that abstraction layer that makes it even easier How is that going to shape the future of SDN? What are your thoughts on that? so what it takes is for a Community effort to pile on and be involved so with p4 org There's a hundred companies that are members of that that that that organization contributing code Modifying the behavior of their networks contributing tools and compilers and things like that for cord There are many operators that are deploying it. So as we have this community growing then There is a there's a little bit of a risk of fragmentation as we all go off in different directions And so finding a balance and we all have a common interest here Which is to create production quality software from odl from onos from cord from p4 from all of these different these different efforts So that we can all take charge of the networks that we own and operate And so it will be it'll come from that need and from that need from the community In an environment where we can create lots of new ideas We've got to be careful that we don't become too attached to those ideas because that's where the fragmentation comes from So we need to find that right balance. So working together in that ecosystem in a meritocritous way where we share openly the ideas that we have I think that's what will give us a path to much more open-source use in our network that we share as an ecosystem All right. So so Chris a question for you as Nick was talking about The you know notion just having more control of the data plane the forwarding plane There's a notion out there that there is this intent-based, you know networking concept And I wanted to ask you how how is intent-based networking? You know really help developers operators to have better control of the the packet forwarding the policy engine and you know, how does that come together? So everything can be solved with another layer of abstraction and intent-based networking is And is intending to create a higher level abstraction so that you can define Your goals in a language that's relatively simple divorce from the underlying implementation details So I think that's it's a very powerful concept like anything else You we always have to kind of distinguish between the excitement and the hype and the reality but policy-based management not even network specific but policy-based management is Absolutely a thrust in the industry because what we're doing is continuing to build more and more complex systems complexity Needs some level of abstraction to help operators manage the complexity Otherwise you're going to drown in in all of the details So intent-based networking is is an example of that you know bring bring the abstraction layer a level higher use something That's declarative describe your intense not describe the details of IP addresses and and ports and Ackles and all the kind of header level details It's an important step but I think the The overall picture here is we're trying to build Next-generation networks at one level. I think we're trying to really simplify the fabric and maybe move complexity into software Where we can manage that complexity A little bit more rapidly because hardware does have some lifecycle limitations And at the same time find enough commonality that we don't create a new set of Complications for ourselves as an industry if you look back at the evolution of networking. We've had a proliferation of standards there's a lot of ways to do similar things and We're going to do something similar in open-source communities will develop a lot of ways to do do similar things And I think what's challenging for us as a broad industry is finding the best of breed ways to do that and and really Accelerating those as core focus areas including the software for the industry so that we don't create fragmentation and You know part of that fragmentation is a lack of interoperability The part of that fragmentation is just focus and we have finite resources across the industry both from an operations side and a developer side and there's attention between innovation and kind of Contraction that brings focus on on a smaller set of projects. So kind of an exciting time for me in that sense, but it is Hard to figure out how to work together as broad Communities across these different projects and find Whether it's intent-based or whether it's you know some some next AI driven autonomous networks You know all these concepts are really exciting but finding the key ones that we can focus on together I think is where the real power and benefit will come from Yeah, I do think that unification is very critical for the networking community to scale and drive even more innovation It's it's interesting. There's a trend of you know the community sort of you know created all these different open-source projects And you know there were a lot of ideas Odl and opn av and Onos and there were a lot of these projects and somehow I think it's coming back together again And there's unification not only at the organizational sort of community level But also at the technology level at some of these abstraction layers are coming together I think this conference has been exciting because there's a lot of talk about AI and You know the convergence of these technologies are kind of coming together Guru I want to ask you you know looking forward You know SDN in the next 10 years If you have your your crystal ball What would you say about you know some of the emerging things that you see today? And how is that going to change the direction of SDN in the next 10 years? As you know in I think in the technology sector making predictions is not a wise thing to do So I think I wouldn't want to give you any predictions of that type, but I think general trends I think that are So I think that are you can look at technology trends and where are the area where they would get applied? I mean I think I have a thing earlier that Edge is definitely Not because of cod but in general I would think that in the service provider industry the edge of the network will definitely go through a major transformation and I think It is pretty clear that when that edge transformation happens We will be leveraging SDN desegregation white boxes open source and all of that and I think that will You know we'll I mean I think all this trend suggests that that is you know We are on a path to doing that and that will happen in that space I'm also I mean I think Chris alluded to this to some extent I think the if you still go back and look at some of the early papers on SDN I personally think that we still have ways to go to realize the full potential of SDN because What was conceived? Even early papers of SDN is that through this logical centralization of the control plane and the configuration You have a global view that represent forwarding state performance state and Configuration state and once you bring all of that state at one look place then there are many many opportunities to do things in terms of network verification network debugging Kind of change management how you do it? The kinds of thing that the intent based networking and so all of this Software stack and tool chain that can be built on top of that logically centralized control plane All that work is still needs to be done whether it is AI plays a role in it and machine learning plays a role in it That is to me is a secondary but all that You can say exploitation of real SDN is still needs to be done And I think it will be done and I want to make this one last thing that if you look at where Google is going I know a lot of times we say Google is very different But Google has already moved quite a bit in terms of logically centralized SDN control plane And on top of that building this tool chain that is helping them get both Agility as well as very high availability of their infrastructure So they talk about being able to do changes maybe once a day or a few times in a week And at the same time making sure nothing is breaking in their network And that is because of the fact that they are exploiting some of these SDN Capabilities and I hope the rest of the industry will do the same going forward great So we have just about three minutes left I want to just ask the panelists. This is probably a question in the audience minds as well we have a community of operators You know open source developers and users, you know people who deploy networks out there any sort of You know recommendations and how they could contribute towards the growth of open networking And you know, how can we collectively shape the direction of networking in the next 10 years? So maybe I'll start with you Nick So, you know this SDN is a bit of a funny name Software to find networking as Vince said it's already defined in software What is it actually about like nearly all disruptions in in networking? This disruption is about moving who is in charge moving who is in control It's moving from the control being inside the box closed and proprietary into your hands as the network owner and operator So that's the change that's actually going on this change only actually happens if at the receiving end There are people who are willing to learn about the open source get involved in the ecosystem Contribute that's how that change actually comes to fruition So there's a handing over of the keys that's taking place and the keys have to be grabbed and that takes a lot of time a lot of commitment a lot of effort and That's what will actually bring this to fruition Whichever project it happens to be and whichever part of the ecosystem that happens to be Kuru Yes, I think I also like the Vince answer to the question that Arpit posed him at the end and for when to say the following That now for this generation of people there is another opportunity to recreate the internet I thought that was the most inspiring thing I have heard from somebody who is the father of internet and so if that doesn't inspire you I don't know what can so I think I really feel that with either with because of SDN because of Disagregation because of open networking for our generation. I'm getting also getting old But some of the near the younger people This is your another opportunity and that comes once in lifetime where you have an opportunity to kind of reinvent the Infrastructure of the society the most important infrastructure of the society because of the changes that are happening So I hope you will grab it and do exciting things with that Chris Starting point. I'll happily make a prediction SDN at 20 will be really really an open source movement so To to next point Getting involved is how you can manage Influencing this direction, so I think SDN is about you know Unlocking the potential of the network in the context of applications and users not just the operators trying to connect two ends of Two different separate end points Getting involved looking at how you change your own Mindset and skills because many of the operational practices that we see today and networks don't translate into kind of a Software world where things move rapidly and we look at Being able to make small consistent incremental changes rather than big bang rollout changes So getting involved and really being open to to new techniques and new tools and new technologies I think is how together we can create the next generation the next the new internet internet Three I guess Sounds good. I guess you know my my last word to everybody is that the community is extremely important And I think all of you for your contributions and regardless of what your role is you may be a developer You might be an operator. You might be You know a tech writer There are many different ways to contribute to to the projects So, you know regardless what your role is and which one of the open source networking Projects you're involved in continue to contribute, you know, we deeply appreciate it you're making the difference and You are essentially shaping the future of SDN in the next ten years So what's that? I want to give everyone a huge round of thanks for being here and being part of the community And let's give a round of applause to our panelists