 All right, we have a packed audience great So good evening everyone my name is Narendra and Iran one of the solution architects for for for red hat We're the focus on telco And I also want to quickly introduce Sanjay and Ian All right good. I'm Ian Hood. I'm the chief architect for the telco vertical and joining my my colleague here Narendra Sanjay Igari senior solutions architect for a telco vertical So again welcome what we're going to be talking about today and is is you know just just a collection of thoughts and ideas you know from working with customers and some of our experiences and It's it's a it's a topic that is actually very close to you know everything that we do today in terms of watching content over over the network and the title we came up with is basically taming the tantrums of choppy jerky laggy video and The the reason we have we have all these we've you know these these terms in here of misbehavior are because that's exactly how or this is exactly how Video sometimes our content delivery sometimes behaves you know, so how many of us in here have a mobile device and You know, we want to play something and we always experience an issue and we experience that annoying little buffering Right, so what we're going to be proposing here is Some of the the thoughts behind why that occurs and how we aim to to address some of those issues So the first thing I want to talk about is you know, we talked to a lot of our customers We talked to folks like yourselves and the first thing that that we encounter is Red Hat and content delivery You know, how do the two relate to each other? How do you you know? What does red hat have to do with with with with this with this topic at all? And why are we even here talking about this? So as you know today that that open source is is ubiquitous, right? And it's and it's all about how we can deliver some everything that you see here these these these terms in terms of having a very malleable flexible architecture it's about how we can address the needs of not only legacy but also how we can evolve into To deliver will actually traditional a lot of the traditional content over new modes of delivery But then also addressing the new formats for this delivery. So things like You know as we evolve from 4k and ultra HD 8k and even things like virtual reality and then it's about Encountering as you deliver these services as you encounter Issues of quality and you know just delivering the best quality of experience to the end user So with that, you know all of us have experienced as I mentioned some you know some of this to some extent which is You know you expect to see the screen on the on the left You know you don't want to sit down with your family and you want to watch a movie on a Saturday night And you expect to see that high definition seamless transport and what we get it often is What you see or observe on the right that little buffering ring Right, and that basically is is quite annoying And it and it's very evident from the statistics that you'll see here, which are based on the the startup time So basically anything less than or anything greater than six seconds of startup Results and and these are these are statistics that have been provided to us You know by a lot of the the providers right the the media providers or the major media providers so You know this isn't just some some point of conjecture or something We've just pulled out of thin air these are actual real statistics and you can see that they're very real because Anything greater than six seconds of startup results to a loss of audience Right, and that's not good because especially when you're trying to deliver content You know trying to monetize that service. That's a problem Right because the less or the more you lose your audience the less subscribers you have to that content and essentially to your service and then There's a there's an exponential increase in That as you experience that delay and it as it goes to greater than 10 seconds You'll see that about 50% of your viewers are lost And then you'll also see that even with a 1% buffering of video You're gonna experience about 5% of loss of your viewers and you know, we typically So they've been characterized or some of the reasons have been characterized In one of these sort of domains, which are you know the whether it's bit rate or whether it's the resolution of video That's being or content. That's being delivered or the frame rate the frame rate or even even picture quality So having established the some of the challenges here And in terms of also addressing, you know, why why we're here today and the topic we want to talk about I'm gonna hand it off to to Ian to talk about why we experienced this So and so I guess, you know, so why does this actually happen? So, you know, it's pretty obvious that you know The dedicated networks that have been built for our homes and everything else were designed for this or at least reasonably well But you know, it wasn't really designed the internet to actually handle the streaming in real time And if I go over the top with video whether it's Hulu sky Netflix any of those I want to be actual to deliver this experience and Unfortunately the ability to do this with standardized and Accepted prioritization mechanisms. They don't actually exist and that's kind of one of the reasons that we're here is to how do we actually take these Things we're talking about and standardize them again take that open source thing to it So the real simple picture you see there is that if everything's running beautifully, right? All things are going, you know add in a very specific rate But what happens is if the delay isn't consistent therefore, we have jitter That's actually much more important now things come at different times and now not only to have Video drops and things of that nature But because video has very specific packets in it an actual control information in it if I lose an important packet I don't just lose like a little pixelation on the corner I lose the whole thing right and then on top of that when that starts to happen Then I get overflow situations and as a result of that now I get the rebuff ring and things are spinning and You know and my son has already left the room He's okay. I'm done with that and that's the other thing is that you know as we've you know look at what people are Extending as acceptable. This is really an important issue for us And so these are the things that actually you know get in the way of what we want to do So what happens when we retransmit pockets? We drop them Then we end up with shared media situations where I'm sharing it you know in in cable environments A real issue for us is as we look at higher definition We'll see in a minute is how long is it take for the actual round trip You know for the data to get back and forth and then what's the delay in things of that nature of how we actually deliver this So this ability to deliver now packets in a very consistent form needs to be standardized because at the moment You know it's a very complex system to actually build and you can't kind of change things on the fly So as I said what happens with real round trip times, so this is a real issue for us So they've actually got it and measured if I decided to try and actually try and deliver you HD If I'm more Now than 20 milliseconds, right? I can't actually deliver 4k you HD so I actually can't do it So since I do that we started seeing this rebuff ring and those kind of things So as we move to this higher definition and virtual reality I've really got to bring that time in which we all know there's different ways to kind of solve this and we'll get to that in a sec The next one up is let's take about how much bandwidth this stuff actually takes to deliver So over the course of time we've gone through you know standard def one two three four and they've slowly You know started to chew up more bandwidth Then we get to sort of the first high def You know and we're kind of in that you know six to eight megabits per second But look at the bottom of the list here is that we end up with 24 just for one stream And this creates a significant You know doubling of requirements both in storage etc And on top of that that issue of the round trip times to deliver and another thing people think that okay Well, why don't I just you know, I'll just ingest it once and they just send a single stream at a time Well, that's nice if it's the lower speed ones because I can get that in fairly quickly But the you HD itself is taken you know more than all of them together anyway So we got to solve this problem either. It's not just for you HD So I'm gonna hand this back to a Narendra and he's gonna talk about okay So let's really see how bad you know This is happening over the course of time in the actual network here in North America, right? so You know and and this pretty much says says it all right So we we've understood or we've got to a point where we understand where where this congestion is is coming from But these are some some actual statistics right over the course of the past two years and I want to just you know walk through this a little bit because You know if you look if you look at the top chart There's a couple of differentiating factors between between what's displayed on top and at the bottom So let's focus on the top for a moment. The top is mainly so they're both They're both distributions of Peak period traffic, but one's for fixed access and one's for mobile access, right? So first differentiator, but while focusing on the top you'll see that it's also there's also a distribution And these are aligned by by year on year traffic, right? So there's there's 2015 through 2016 so you can clearly see This distribution as as we're moving forward in time and then also share with you some interesting results that we were looking at this morning as we as we move into 2020 and actually you can you can actually obtain that from from some public domain information, but This is upstream traffic in the first two columns and then you can see the downstream traffic So this is actually you can clearly see how it's exhibited here, right in that in that big red bar Where it's it's demonstrated that a lot of that traffic is us, you know consuming that content, right? It's it's the stuff we're watching, you know OTT over the top. It's it's all about the OTT services. It's the Netflix the Houlou's It's all the other content that we're downloading from and watching, you know via YouTube and Then you can see all the other types of traffic whether it's communication or gaming or web browsing And you can see you know the internet what what started off You know with with everyone consuming it for web browsing you can see what a small portion or slice that's actually being consumed Just for web browsing. It's almost negligible So this is part of the problem, right? And then when I when you talk about fixed access and then you look at mobile access You can see the same phenomenon, right over here as well that this is the type of upstream traffic And you can see even the upstream traffic, right even on mobile devices This is the types of content where we're recording on our phones as we're moving around and we're mobile and we're uploading it So, you know, there was some interesting studies that that We found in some some Ericsson studies where the volume of upload traffic is is Gradually or almost to the point where it's going to overtake the amount of or the volume of downstream Traffic and and that's a huge problem, right? So not only do the devices have to keep up, which is what we're seeing But it's also the the the traffic or the volume of traffic as an aggregate that's going to grow And then I'll let Ian quickly share some of the statistics we found this morning as we move into 2020 This morning was that I've looked at some of these studies a few times and Effectively the traffic that we have monthly, you know is in the exabytes So that's like 10 to the 18 kind of numbers on a monthly basis And if you took, you know the sort of, you know, 40 million of us, you know All day all the time for an entire year, you know downloading a you HD video Simultaneously, so took you know a pretty large population. That's how much traffic is actually going, you know monthly here in North America And you know, I'm looking at these charts and one things that I'm sort of expecting is that a The size of these graphs in aggregate would be larger And as you said Narendra the key thing that's going to start to happen is that that upstream numbers actually going to come up As we see that our our kids are sharing more things and you know Periscope is in and all these ways that I can do FaceTime live from from anywhere. This is 2016 data So that's actually a little dated already, right? So that's the thing about you know, you look at data You got to project where it's going to go and so that's kind of the thing We're seeing there now with the data that's out there And just one last thing I want to quickly share here before I hand it off to Sanjay here is You know Netflix who's one of the premier providers as we all know of content, but also high definition content agrees right so just you know There were some there were some Comments made by Reed Hastings CEO of Netflix, you know at Mobileball Congress This year and then also last year at I believe it was at at CES and Some of you know, these are some of the some of the comments You know, you can just read through them. I'm not going to read through them But the idea is that you know, you can actually go to those links You know the presentation will be shared but you can watch them and you can and he talks about Specifically what we shared in our first slide, which is even Netflix has not been able to get rid of the buffering problem Right the premier content provider right OTT provider has been not been able to address the buffering problem They will at some point right and you know just along with with the rest of the of the providers But they have not been able to address it and there are some severe challenges as to why and some of those will be you know We'll be discussed here. So with that You know, I just want to talk about the Some of the some of the challenges, right? So we know why it's happening, but these are some of the the challenges that are being faced by these by these providers And I've sort of split them up into this this multi-dimensional Space because I want to I want to be able to separate them out The first of course is you know, we always attribute things to to technical challenges, right? But they're not always technical There's also issues with with operations and there's also issues with economics, right? Does it make sense is the market is the timing for delivering this type of content and the geography in which I'm distributing the content? Is it suitable right? Is it the right time for me to release this release the service if I release this service? How do I scale if I scale it will my will my supply of my services meet the demand of my customers? Will the customer demand even exist a year from now, right at the pace at which technology and Delivery services are being are evolving So very quickly You know, it's about content licensing So it's about it's about creating this equilibrium, right of and that and that balance of infrastructure and sort of offsetting the the investment with the The cost of well, what's coming in right? I'm charging for the service. I'm gonna off be able to offset it offset it with My subscription costs and my ad revenue The new services what type of new services can I offer? and then how can I create a model whereby that's justifiable where the declining costs of hardware can be Factored into the cost of my solution and then time to market as I mentioned, right? Is the time for releasing the service, right? Is it ripe for me to go out and grab these subscribers, right? How much market share can I actually gain? And then things like regulatory, right? So some of you in this room who are here are familiar with some of the regulatory issues and about shared content and how that content can be You know the governance of that content right due to DRM to digital right management issues And then also the type of content and where it's stored and whether it's a shared versus a unique type of copy But at the bottom are more some of the the attributes that are closer to our hearts, right? The operational aspect of things and the technical aspect of things which are you know when I do the start to deliver this at scale I need a degree of automation. I need a high degree of automation, right? So gone are the days when I have You know for each asset on my data center floor or my infrastructure I start to manage this with with as many people right that model simply doesn't exist and Then how do I upgrade it and how do I evolve it over time? And then I need to be able to collect all the analytics from this infrastructure so I can make it more efficient And I also need to evolve the skill sets and and be able to train my my force my workforce to be able to keep up with This type of architecture and then be able to monitor and remediate the problems and be able to troubleshoot them Very very rapidly and then last but not least some of the technical issues right the scale and elasticity of the services the actual infrastructure The performance the bandwidth and the quality of experience because the quality of experience is going to decide how much How you get to keep your customers if you get to keep your customers and how much you can charge for your service and Then also being able to accommodate the the capacity of this content or the storage capacity of this content very very efficiently And then more efficient use for example of the radio network So being able to at the edge being able to deploy certain services, which can take it take or make or leverage some of the Services at the edge to make your content delivery and your cell towers more or the use of those cells more efficient so with that I'm going to hand it over to Sanjay to discuss and Talk about some of the traditional versus the new architectures Thank you. So when we look at the Traditional way that content was delivered the question becomes That looks good. Okay. I have a content provider. I distribute it to a cache I go over a telco network and the customer has Set up box that receives it. So what's the problem with this? Well, all of this depends on three things you know what content you're delivering you know when you're delivering it and You know where you're delivering it, right? You know that it's you know what program You know you have a file for that You know that you're sending it at a certain time and you know it's going to a certain set of Residential houses at you know in certain neighborhoods. Well once you cut the cord All that is gone. You have no no longer any assumptions What content well people consume any of You know Millions of programs of streaming wait when any time, right? And maybe even consume it in chunks at different times and where mobile right? I mean it's on your phone. So You know where you're locating the storage has to be very flexible It can't be only in in one place in fact, you know the residents may not be The best place to store it, right? Or near the residents may not be the best place to store it may be Minority of the content consumption. That's what's starting to happen. So All the assumptions made in the building of these networks are turning out to be old-fashioned So we need to change so so when we so when we look at The results of this we have to move from this type of environment where Everything is sort of sort of centralized and You know to a more modern environment So what does that mean? Well, we need to have some data around how well the video is being delivered what kind of device it is and You know what happens when you pause and rewind and all those types of things and these are challenges With a streaming environment. So the the first thing first thing you want to do is to actually You know relocate Your caches in different parts of the network. It's not it's not centralized and you see this with the main Content providers you see, you know Netflix Amazon Google. They're they're all doing this but Today It's Essentially a siloed stack So if you're not one of those providers if you're one of those providers you have your own servers you have your own content caches And even content cache hardware that you install And you have your own clients that you write for every Client device out there whether it's everything from mobile phones to TVs to whatever So if you're in if you have that silo if you invest in that much you can actually do this But if you're not one of those providers and let's say you have a license to a lot of great content Well, guess what you may not be good at those other Elements and then you're not going to be able to deliver the quality of experience. So And that's that's a really Difficult problem. Why is it difficult? Well, I'll tell you that, you know, the NBC app on the Roku like constantly crashes Whereas Netflix doesn't so, you know, if you want to watch Saturday Night Live, I'm sorry. I mean It's not a good experience and that was my experience right because you know, they're not they're not a software provider They're you know, they don't that's not their focus, right? So so these are these are the kinds of these are the kinds of things that You know, we would like to get to a more horizontal environment where there are standards and content providers can deliver their stuff content caches can do their thing and and users at the edge can use the clients of their choice, but that means these communication mechanisms can't be Proprietary so the other the other things that that we need to do And in this kind of environment is to Do some sort of have some sort of intelligence in terms of What's being like what type of device is being consumed on to transcode Transcode and deliver the resolution that's appropriate for the end user. So you're not delivering Ultra HD when the device can't consume it That makes that creates a lot of control logic and control paths that go back and forth that need to be negotiated So this is a a interesting topic so As we as we move to more mobile edge computing environment We have more options to move content out towards towards the edge So When we get so some of the benefits of doing this are that you know We can massively reduce the bandwidth and solve some of these some of these problems that we've that we've talked about And we're gonna talk about now Some details around the Orchestration mechanisms which are necessary to make this happen because You know, it's not fixed anymore as I said if you have to if you're doing place shifting and you're watching on one device You want to continue watching on another device? Well, unless Unless again, you're playing in the same ecosystem if if the new device has say a generic app There better be some standards for the handoff. So with that, I'll Pass it over to Ian to talk about some of the capabilities here. Yeah, so the Also some things that we can actually you know do since we've got sort of some additional capabilities from different vendors that come up with you know Software flexibilities to deliver this now stuff at the edge We still can't forget even you know the networks themselves have to be built such that I can actually deliver this Multicast traffic even though it's gonna be we all think it's gonna be unicast forever I'm gonna want to have you know that kind of capability to optimize that bandwidth We want to take care of this round trip Time and we want to actually make it so that you know, we all know that you know I'd like to watch this my son like to watch that my daughter like to watch this But even as a as a family or as a group you want to kind of pick and personalize your Your content so if I haven't got a software way to kind of do that from the portal and actually choose that content that becomes an issue for us and You know gone to the days when I can say that okay I'm gonna lock up these radio channels on the on the broadcast line for cable and that's all I've got to work with right We want to be able to actually know how much I need for this as Sanjay said You know taking the transcode off of this device I've got to move it from my TV to my you know my iPad or to my mobile device And then I got to be able to walk out the door and not have it actually You know stop in the middle of it because I want to be able to take this and and see it there so that's where we want to have that flexibility for those handoffs and Truly be able to kind of take care of this not just for the content But the control channels that are going on In the radio networks because they're gonna be on Wi-Fi for a bit switch to another Wi-Fi Not same dice different or sorry different device Same Wi-Fi. I'm gonna switch over to LT as I leave. Well, how does that actually get negotiated and handed off? We see that all the time right now, you know for your meetings and we're trying to do that You know the meeting cuts off and I've got to come back in And we've got to make sure this is truly modular because it really is all about economics, right? I can't really do this now We've seen with the set tops that you know the truck roll to go swap your set top because that's what goes on Right now is there's something goes you know goes bump the code's not working What do you do you go down to the local store and you know handing back your set top box? You got another one well They want to be able to change this stuff on the fly and make it you know economically viable for themselves That's really the big cost is their ability to deliver that and we got to take that you know technology We've got you know sort of in the cloud side and push it as far as we can out to the edge of network Because that's where really all the costs are so with that I'm going to pass it back to you my friend and We'll talk about where we're going next. Yeah, sure. So so one of the things that one of the things that the You know really good content providers do today is they they have a mechanism to feed back information from their clients as to how well The video is being delivered so that then they can make adjustments on on the video delivery So for example if you turn on I mean there there's some if you get It's a nice experiment. You can do you get find an old DVD player You know from the early days of Netflix integration of DVD and those apps had Debug modes that you can turn on you can see like as the video streaming. What's the Rate and and all that and I have one of these I have an old Panasonic and Netflix is really good at making sure all of their apps going all the way back They always work so you know, they don't deprecate them so And you can see that they'll do some interesting tricks like when they start streaming They'll start at a lower resolution because they can get it to you quickly because that's six second mark They know that as well as everybody knows that they're just what are we going to do about we're going to make sure it starts Before that six second mark. So you know, so it'd be a lower resolution. We guess what that's like the credits or whatever, right? So and then then they refine then that's once a real once the slow start in TCP gets to the right point then they Switch and they switch to a higher bandwidth and and you get a better resolution and things start You know, they're reasonably good. Once you once you get to the once you get to the Program going so, you know very quickly you learn that no don't fast-forward five minutes into it to skip the credits because you know You actually want slow start to Get through get through the point. So they do a pretty good job, but again, it's proprietary, right? So how do we bring that to a more open environment where you know, it would be great if if we took advantage of You know, some of the open technologies we have to drive Some standards in terms of you know, what are the messages for? bandwidth Bit rate and resolution shifting that should happen, right? I mean nobody's really driving those kind of standards and and I think Because red hat has open Interfaces, you know, not only just open source, but we really stand behind Open interfaces between components and we try to participate in that kind of environment This is an area where There's a lot of value in having these open interfaces and where they are not there today. So it's a it's essentially Sort of a in some ways a new market for open stack to be involved in And really promote that openness So if we if we look at the other the other thing that's really important is is that we Do this in a way that's very efficient because there's not a lot of processing left over after you're doing transcoding and all of these things that have some complex integration software, right? so with that I'll talk about I'll talk about a so-called overall architecture You know the NFVI Platform to deliver the networking functions obviously Running on open stack. There are a lot of other control operations Which we'll talk about but one of the things I wanted to really focus in on which is which would be sort of a new thing is Some basic functions that are needed that need to be standardized in order to make sort of Open video successful Very simple bit rate shifting needs to be a standard way of bit rate shifting and we work on that in the industry Pause rewind fast forward most basic things. This is an area where we've over the last 30 years We've gone backwards in technology if you remember the VCR you hit rewind Instantly you can if you miss the joke you get it again now DVDs you hit rewind it took you like, okay It was very blocky then you finally get back and you probably have to re-watch the last minute of content to get to where you missed Streaming you hit rewind. Oh Buffering you're gonna be with you know very quickly. You know, you know what? I'm not I'm never gonna rewind and there, you know, it's like back to the days of TV without VCRs or anything like everything's live Okay, you you don't get to see anything again. Sorry because the buffering is so bad if you rewind So what we've really got to do is you know, if there was just some control message Just I could go back to the content cache that said hey rewind was hit just move your pointer back You know everything would be fine You don't need to re-buffer from the beginning of the stream and then do a seek again and then find it And then find the keyframe. That's what they're doing today so this is this would be a really interesting mechanism to feedback and you know There is a lot of work going on today in terms of In terms of Real-time control we're doing this in many other areas. We could do this in video As well the other area is is of course the telemetry data from the end the clients Why does that have to be proprietary an open client should be able to report its? frame rate with an open mechanism, so with that I just want to take a brief. Yeah, you want to talk about Couple a couple of points there So now as we looked at this architecture now We know we got to push things out to the edge to try to you know optimize things So that's that's one piece of it But in addition to that everything's been sort of been done at the head end Which is you know, which is good and I kind of move things around there And we've got to look at that. You know, how do I deliver it to the edge? So as we take a look at you know that you know those things I want to do out there I'm constrained you know in the actual Radio networks themselves right because we're you know, we're saying this is all going mobile. So there's a number of different Vendors that are actually looking at how do I optimize not just you know the traffic in the head and because that's how we've done it for years I've had all kinds of different technologies that sit behind the packet core that optimize TCP buffering and do video optimization And that's kind of the things we've been doing right now We've got actually ways that I can do this not just in the IP streams Because if I do it there that doesn't actually solve the problem for the actual radio waves are out there that we're actually trying to utilize So there are vendors out there that saying okay Why don't I take a look right inside those and see what's going on there and optimize? streaming content video at the very edge of the network and depending on who you talk to it's a Variation on some people call that cloud ran and some people call that cloud air because I'm actually optimizing it out to the air interfaces themselves But there is again a software module that is going to be open source and made available to us that I can actually do some of these Things that Sanjay was talking about so that we can't have the experience we want Anywhere that we tap happen to take it so it's really kind of taking that Functionally we had kind of out here centrally and found that it didn't really scale didn't distribute didn't get it down to the narrow level That I wanted to it and have that granularity for everybody especially since I'm changing Devices on the fly networks on the flies not just the video content It's actually the control information that does the handoffs between the network So we expect just to seamlessly go from here to there and not miss a bit. That's kind of what we expect So back to you and render almost out of time it is really want to quickly tied us back into the four principles that I brought up there for architecture which are economic and Technical as well as operational and you can see what's demonstrated here, right? This is not this is more than just a pretty diagram with some colors in here, right? This is all about customer feedback and this is about what we've built for our customers, right? So as you'll see here, you know all the way from the bottom of this stack from the NFVI architecture from from the NFVI platform All the way to the telemetry. This is all about optimization and feeding back into the system So you can allow and afford your customers that level of automation and operational efficiency So again with that if there's any questions and answers I mean questions and we'll be glad to answer them but if not Thank you for attending and Let's go ahead get some drinks Any questions any questions if you can Kyle actually no Kyle go ahead go ahead Yeah, so let me repeat the question. So as you're handing off Netflix from one place to another Yeah, or any any provider really this example you're handing off from your fixed broadband to mobile Where does the information get stored that allows you to do the handoff and really there are multiple layers of handoff? because sure the network handoff the network handoff happens when you When you can't reach one network and you start going over another network the problem is To start streaming on another network and to have the video itself be delivered as opposed to just the packet stream be delivered the video Needs needs to be a lot more bandwidth needs to be available and a lot more content A lot more data needs to be available to show an entire frame. So what happens is you typically get a gap So to solve that there are a number of ways that have been solved that have solved that I mean one way is Just always keeping a certain amount of buffer in the client so that When that buffer drains by the time the buffer drains you you're on the new network And you assume that you can hand off right? That's a simplistic way, but more and more you're starting to see More intelligent solutions like if the performance or we're like keeping track of multiple networks at the same time If the performance of one starting to degrade start ramping up the other one Things like that. So there are a lot of different Ways in which that's being done. It's already being done successfully in phone calls, so for example AT&T worked with Apple on the Wi-Fi calling and That that works pretty well. I know because I have really bad LTE service in my house. So it's almost always shows the Wi-Fi indicator And you know, that's and it works really well. So Can we get to that for video? Well To get to it for video is going to be tougher because it's not just Two companies working together, you know, big service provider Big device manufacturer and saying, okay, well, we'll make it work between us and I'll cover almost everyone It's really going to have to be an industry effort With multiple multiple players at each level at each layer. So but I think it's something worth doing any other questions