 All right guys, thank you guys for coming to open control days for that open source so Today we have a lot of good. We have a good lineup and to get things kicked off We have Randy buys from Juniper Networks the VP of technology and strategy. So let's give a hand for Randy Hi So fair warning, I'm Sick and sleep-deprived. I made the mistake or I don't know. Maybe it was a mistake of bringing my Family with me because I was doing a two-week trip and so they my wife and my 17-month-old baby got sick on Saturday And then I got sick yesterday And so she woke up, you know screaming at 3 a.m. This morning for like a half an hour before I got back to bed So I'll do my best. It might not be in my a game, but I promise that at least what I have talked about will be interesting Sound good. All right, so let's go So it's just a quick overview of kind of arc. We're gonna talk about today I am trying to tee up the rest of the day So that some of the presentations will be in context and I'm also trying to tee up the open control users group tonight Where we're gonna talk about some really important things about how we're growing and leading the open control community Which I think has been a source of angst for quite a few people so You know what I should have asked this at the beginning. How many people know who I am raise your hand Okay, so most of you so just a real quick introduction for those who don't My company cloud scaling was one of the founding companies for open stack in 2010 and Then I built it up and I had two marquee customers AT&T and Walmart Both of whom had 25 rack deployments And then I sold it off to a company called EMC which you probably have heard of so I've been in the open stack thing from the very beginning and then before that I've got a long storied history Around infrastructure and cloud to give you another example, and I'm trying to not give you my whole biography Which you can find very easily with Google But to give you another example in December of 2006 when I was working on my first startup We were using this open source tool that you probably have heard of called puppet and at that time puppet community was about 30 people There was no puppet labs Luke was living in Tennessee and working from his home And we build an orchestration system around puppets so that you can deploy an interior web app in a DevOps style onto Little system called the Amazon web services elastic compute cloud which at that time was in beta Private beta and you could order a single size VM for 10 sets an hour, and that's it No EBS no load balancing no route 53 Like none of that right so I've been doing this for a long time And cloud is my specialty, and if you Google for me you'll get a ridiculous amount of information that you don't want to know so Why that's interesting and why I wanted to talk about before I go into this slide is that In the early days There was a company that was a second company of public cloud that I was a CTO for called Gogrid And I started engaging around SDN at that time with a company called Nasera That was in 2008 right so South Florida find networking hadn't been coined as a term DevOps hadn't been coined as a term We were just looking at these at these overlays and saying, you know, this is really interesting technology What can we do with it? And for me, you know, I was just always skeptical All right, what am I gonna do with this overlay right it's adding a lot of overhead, you know It's maybe gonna slow things down. I don't know if I'm gonna get as much visibility into what is going on at the physical layer What do I want to do when I troubleshoot? That was my primary feedback to Martin Casado and and Steve and the Nasera team as I said look You know if we put this in place like suddenly like a lot of the tools that I have for troubleshooting network problems Like don't work or don't work very well because it can't make connections between what's happening in the Virtualized abstraction the overlay and what's happening at the physical layer like that's really important I need all that tooling need the operational pieces so You know, I've been skeptical about overlays for a long time and now I work in the Contro business unit of Juniper, right and part of that and part of why I started to get comfortable with it is that sometimes Technologies have a time and a place and people deploy new technologies and they're just not really ready And so the realization I came to over the last year is that you know There's actually a killer app for SDN overlays, right? It's like virtualization the first time I saw VM where I said why would anybody do this? Like I don't need to virtualize You know and put multiple servers on a on a x86 platform And then I had my aha moment eventually what the killer app was for VMware. That's another story And but the point was is that what I've seen today in the past year is sort of the aha moment And I and you'll get a taste of that today when riot games presents later. I think at 3 o'clock 3 30 330 So riot games and I don't want to spoil their thunder But what they did is they said look we've got this great abstraction called compute and we've got this great abstraction called open Contro or sorry against I'm sick this great abstraction compute abstraction called containers I've got this great network abstraction called open Contro And if we marry these things together then suddenly, you know our developers can write to a single kind of Framework as a common system common abstraction and then whether that is deployed on top of public cloud or private cloud nobody cares and So I think that if we look at SDN overlays and the complexity they add in both deployment and management troubleshooting Suddenly having them as sort of a network abstraction is sort of a killer app So you can imagine something like this You can run overlays everywhere. You don't care. It's in a classic legacy bare metal data center You could run them on top of public cloud, right? Why use Amazon web services VPC if it's significantly different from Azure's if it's significantly different from Google's instead You could have the same security and network semantics across all the public clouds Oh and also in your private cloud whether it's VMware or open stack or just legacy bare metal systems And what that allows you to do is when you've got the same network and security semantics everywhere Is that it means that applications suddenly can be much more portable? Maybe this doesn't solve all portability problems, but it solves a lot of them, right? And so this is why I think the SDN overlays are about to come of age And I think that I want to paint you a picture of why that's really important. So the title of this topic is You know of this presentation is you know, can we build a global ubiquitous network fabric? Well, if overlays can run everywhere and give us sort of a common, you know System for having networking regardless of what the infrastructure is Regardless of what the underlying network architecture is then really interesting things happen, right? Means we can suddenly deploy them everywhere globally, right? We can make them ubiquitous We can make sure they run on top of anything, right a bit like TCP IP mean some of you are probably old hand old network Can't anybody remember IPX? Right, right. Where's IPX now, right other than the US military, right? I mean, there's a tremendous amount of value in having the same type of Systems the same types of protocols and standards that are deployed everywhere Suddenly if you're trying to build things on top of that common set of standards or protocols It gets a lot easier, right? And so I think that that's the picture. I want to paint right what if SDN overlays are not just this thing That's in your cloud. What if there are a thing that's basically wrapping the entire global network and giving you the freedom to run an overlay? Basically anywhere you want right to create your own networks anywhere you want to manage them any way you want Another way to think about this is if you I don't know how many of you are in the carrier space But they've got this thing called MV. I'm new to like all the carrier stuff So I've been learning all this stuff and they've got this thing called a MVNO, what's the acronym stand for somebody who's Mobile Virtual Network Operator so what that means is somebody like Walmart can actually have a wireless network And they're writing on top of all the carriers other wireless networks in a virtualized manner, right? That's that's pretty awesome Google Fi is basically an MVNO that operates globally across multiple carriers Just imagine if you can do that you suddenly in your business can create the opportunity to have your networks running anywhere And you simply do not care where that place is you don't care who's running it You don't care the network topologies. You're they're using the switches the routers You don't care about any of that stuff, right? Your network and security basically can ride anywhere globally So we get application portability, but the side effect of that is that we get to disintermediate the specific infrastructure We don't care if it's FEMA or we don't care if it's open-stack. We don't care if it's Amazon We don't care if it's Google or Azure Right, that's really important because right now we tend to lock ourselves into these public cloud ecosystems Right, they're big walled gardens. They're very proprietary I'm sure you know, maybe the network piece isn't as proprietary as something like Dynamo DB But it's still different right the way that Amazon does security and networking is different from Azure and from Google and from your VMware deployments The other thing about this, you know, which I'd said a little bit earlier is that it's foundational, right? So if we've got this Thing where we've got overlays everywhere. We can start to build upon them, right? We can start to add Capabilities at the edge of each of these overlay networks. Maybe it's SD-WAN Maybe it's a single pane of glass where we can see our security posture and globally so on But it means that like if we can get it everywhere then we can start to build additional value on top of it So what I'm trying to say is that Juniper would like to play the game where open-contrail is adopted by everyone whether they pay us for that or not and Play for the value on top of that Like where we want to start thinking about how we monetize open-contrail is different than the way it's been in the past We want to start thinking about getting it in more people's hands making it easier to use and instead play for adding value up the stack So I apologize that this is a very long list But I wanted to sort of say hey, where is open-contrail today? Where is it with the new 4.0 release and what can it be in the future? so This is the set of requirements that I think are our this is what we need to create a globally a Global ubiquitous network fabric right needs to be an open code in a vibrant community Which we're going to talk about some more needs to run everywhere right public clouds private clouds on multiple clouds and at the edge It needs to not care what kind of computer it's attached to it needs not care What kind of infrastructure it's attached to it needs to be easy to use and deploy possibly one of our greatest weaknesses It needs to have strong transparent community governance and direction It needs to be software to find needs to be highly scalable in production grade and needs to be technically excellent I think we're good on the last three So if we look at where open-contrail has been you Know historically it's been very strong on the technical capabilities you can read these yourselves whether you agree with them or not I just want to put something up there that kind of painted the landscape And it's been very good at being sort of like agnostic to the infrastructure. It's running on But it hasn't been so great to use deploy or manage. I mean, it's a network stack. It's very complicated I remember when we were integrating contrailing to the cloud scaling product at before GMB or even bought them And my team came back to me and they said there's more moving parts here than there is an open stack Of course, that's less true now. There was a while ago There are many more moving parts in open stack these days But the point was is that a networking stack is complicated. It just is inherently complicated That's no excuse not to make it easier to manage and deploy But it is complicated and then the place where we have, you know, maybe we're technically excellent Maybe we've been good stewards of our own project But one of the things that we definitely haven't done Is figure out how to leverage the community and how to invest into the community and how to really be much more of a community-driven project So We want to fix that I'm gonna let you read this In order to Fix that going forward with open contrail You know, we need to kind of Reset we need to rethink where we were like I don't want to make an incremental change I want to I want to make a revolutionary change I want to leapfrog where we could go if we were to just make small changes And really start moving in a in a different direction And part of that means having kind of a new vision. This is something I put forward. That's just a Beginning of that, right? I mean, I think that any kind of vision that needs to be Needs to come from the community as a whole But I'm trying to put strawmen down that people can respond to I'm trying to You know start us In a certain direction I think You know if this is resonating with you a little bit then the first thing you're going to say is like so so what does this mean? Well We have to reboot the open contrail community I want to say reboot reboot. I mean, you know control all delete reset wipe the disk drive boop Right, you know reload the operating system reboot Right It needs to not be juniper rebooting the community. It needs to be the community rebooting the community And the only way to do that is for juniper to give up a lot of the control And I think that's okay and tonight at the open contrail users group We're gonna we're gonna talk about this in a lot more detail And I'm going to put out a call for people to come who want to participate and be part of this you can be Uh somebody who uses contrail today. You can be somebody who wants to develop and contribute to contrail tomorrow You can be a competitor. You can be a major customer Like it all all things are open. I remember You know people came to me for the open contrail day today and they said Oh, you know, marantus wants to present or marantus wants to be a part of that. Is that is it okay? And I said, yes, of course Like that should no longer be a question right So the other piece here is that we want to move towards more agnosticism, right Contra already does this better than all of its competitors, right? We don't really care what the underlying system is at least philosophically But if you look at the actual sort of like Code base if you look at the documentation a lot of it leans towards open stack Open stack is fine. We're gonna all I mean the carriers are going to use open stack till the end of time as far as I can tell And juniper loves carriers So we're going to make sure that contrail always works for the carriers For open stack, but you know if you look at the contrail four died over these We have native support for kubernetes and now we support active directory for authentication I mean we want to move in a direction If if we truly want open contrail in the hands of everybody For every infrastructure and every kind of compute then we have to move towards more agnosticism We have to be less centric around open stack. We need to you know, continue to level up in that regard And you know the outcome of this is that we just want to drive greater adoption and like I said, we are going to play less for People paying us for commercial contrail support and more for the value add that we can provide on top of open contrail So that's that's quite a bit of difference and it's going to Be hard for people to understand at first, but here's another way to think about it Would I rather have a hundred times more people using open contrail and have it be more ubiquitously deployed? Even if I only get 10 of it Yes, absolutely Right, so that's the game that juniper's playing now. I'm just being forthright about it Because what I want is I want people who maybe to date have been a little frustrated with how the community is managed And how juniper's kind of had to be their project to really come back and sort of reevaluate like we want to play with you again Right, I gave um the uh pitch, you know a a 30 slide kind of more detailed plan that I don't have complete buyoff on yet I have you know directional buyoff from from the ceo, but I don't have complete Sign off on yet to a very large carrier I'm one of our largest contrail customers yesterday And they went from being frustrated with us to saying this is what we wanted to see This is what we wanted you to do and we're very excited about this. What can we do to help you? All right So, you know, this isn't a game You know, I'm very serious about contrail being a more open community a healthy and vibrant open source community And we'd like people to come to the fold. So if you come to the open contrail you use a group tonight I'll give you more details about how you get involved Okay And then finally, you know, we just sdn overlays are tough. They're complicated. Um, I think it's going to take a group Of people and a community to basically make them easier to deploy to man to manage to increase the equality To have better testing and test coverage. Um, and so we need the community for for more than just, you know Engaging and driving this thing ubiquitously. We needed to increase the quality of the overall project And that is it Any questions Any quick questions Okay, somebody tell me I'm full of shit. So if you guys have any questions, please get it into the mics Over there, uh, so we can record them Yes, please you mentioned that Juniper's way to monetize will be to do something on top. Can you kind of elaborate on that a little bit? Um, we don't have a very clear picture of what that is yet But we're looking at things like sdn. We're looking at things like advanced security features We're looking at, you know, sort of, you know, looking at open contrail sort of foundational piece Which we will offer commercial support for if people want it But we'd like our tools to ride on top of open contrail or contrail like we don't we don't care And I don't necessarily want to tip my hand on all the things we're thinking about for obvious reasons But the intention is that think of it as moving more towards something like an open core model where the foundational open core pieces of functionality are all open But we'll have our own commercial bits on top of it One example right now would be out formics the new acquisition that we made the dust telemetry and monitoring And it monitors contrail open contrail. It doesn't care, but it's commercial bits For federal partners, I'm kind of short, so I'm not sure you guys can hear me Okay As one of your federal partners that is very much partnering with you in this area We run into a problem with People saying yeah contrail, but we really want to push boxes So when you're talking to the ceo level How is that Being addressed one of the ways we're trying to address it is Value on top with network service orchestration that combines orchestrating this overlay with orchestrating your legacy systems Is that being talked about at the highest levels? Yeah, absolutely. Um, so Okay, the short answer I tried to a chart is that when cloud scaling was acquired by emc My mission became not running in the cloud scaling team but helping emc transform from a hardware company to a software company till del bodice okay, uh, so my mission to juniper is very similar which is that Rami knows that we need to over the long term become less of a hardware centered company and more of a software centered company And if you look at a lot of the value on the juniper boxes Historically themselves like juno s it is all software And so, you know the the the business unit I'm in which is called the contrail business unit Which I hope gets renamed to the cloud software business unit is software only at the moment and there are whole software teams Run by people like alex that are our sales teams that are focused on selling software only and not on boxes And so we're we'll get there the reality though is that you know, you've got this go to market motion That's 20 years old And so people a lot of people in the sales organization are just used to shipping a box But there are new parts of the business like the contrail business unit that don't get any compensation on selling boxes Don't care about selling boxes and are focused exclusively on selling software So we know that's where the future is we're making the moves to get there But it will be a slow incremental progress on that just because you can't you can't surprise everybody in that manner by like, you know Making a sudden drastic change Partners that believe in what you're doing and want to support you. How do we get to you guys? Well, I'm super easy to find It's our bias or randy be a juniper net and then I'll connect you with people like alex and his team Which are totally focused on that and alex maybe can follow up with you afterwards Um, and but you know, you can look at me as a conduit for that if anybody else needs to do that as well Thank you. Sorry my contact information's not here. This was You know, 11th hour last night at 11 p.m. While I was sniffling and You know coffee and hacking I have more questions. I only got two I guess I'm supposed to be off soon Oh, not you Easy question So my name is martin and I've been we've been using open contrails since it was open source basically so But one challenge that I think some people have is Doing the or at least it used to be a problem. So update me if I'm wrong going the juniper distro Is not applicable to everyone because a lot of people it's not what applicable like a lot of people want to build their own You know kind of design Yeah, sure their open stack so Are you already addressing how to support people who have gone that route to basically the run open contrail somehow? Well, we're not gonna we're not gonna offer support for other people's distros if you build your own distro You're on your own, right? I mean what we you know the thing about and I wrote a blog post about this on my blog cloudscaling.com The the thing about um What we're gonna offer support for is there's what we're gonna offer support for is always sort of like a pretested vetted You know structure like a product, right? I mean commercial contrails a product if you're going to take commercial contrail And you do whatever you want to it. It's no longer a product It's like saying can I get support for hardware appliance that I've basically logged into and installed my own software on No, because the very first time that we have an incident I have to tell you go put the original software on you like you take all your You know bastardization off because that's a variability that I can't I can't account for when I'm trying to troubleshoot your problem So but in terms of the community and how we will encourage people to build their own distros Hey party on build your own distro sell commercial support for it Whatever you want and in fact, I hope that be in the next week You will see us open up a lot of stuff that we had kept closed Um, I think there's a good chance I'll get it done in the next week or so And that will include all the build release process and and tooling for building packages as well as all the pieces for building The new containers And so if you want to go, you know take that stuff and use it to build your own distro or whatever that's fine So there still seems to be like a small gap that there are people using open contrail That would like to have like someone who can help them basically like sure Well, that's a that's a that's a that's a business opportunity for another business juniper sells products And so, you know, that's what we're going to continue to do Yeah, I mean it's really hard to have a different business model So, um For your background having worked in the open stack foundation and been a part of the open stack community for a long time What kind of lessons learned from that community? Will you use for the vision of the new community? I love it. That was an excellent question. So, um, I don't want to get into it too much But the way that I view kind of the difference between The contrail community and open stack community is that the open stack community has been very very bottoms up And very little top-down command and control and that's created its own set of problems And junipers have been been very top down And very little bottoms up and so we'd like to chart a middle course and what I mean by that Is that I would like to see, um, us fostering more people contributing to the code base Um, because they're trying to get their pieces done, but continue to have some strong Technical oversight and governance. So it's not, you know, 100 cooks all cooking the same soup And I think that the way that that is to be done is still tbd One of the things that you'll see about this rebooting of the community is that We just don't have all the answers. We actually need everybody to come to the table and talk about it But I've engaged the linux foundation I've engaged the cloud native computing foundation And I've been talking to people like sam ramji who was executive director for the cloud foundry foundation And I'm trying to get more people around table to help us figure it out, right? Because like There are so many different ways to build an open source community. It's not even funny And it's been done lots of times before and I just simply don't see any reason to make mistakes that we don't have to So we're going to get more people around table who can help us think about it And then the whole community has to sort of figure this out But what I would very much like to see is to drive a middle course so that there is strong technical oversight And that we focus more on like itf style running code and rough consensus to get stuff in But we don't just allow any check-in willy-nilly like it still has to it's the network has to be high quality It needs to be production grade. It needs to be scalable. These are all things that we're good at today I want to be even better at in the future And you know when we set the values for the community I hope that we are thinking about those things and it's not just like What's the willy-nilly features I can throw into the bucket because I don't think that helps anybody Thank you I'm way over my time I should probably wrap it up so you guys can get a little break before the next session But I hope to see you all at the open control users group tonight Thank you ready