Cloud Gaming: Where Are We Now, Where Are We Going?

hello and welcome to this GBC talk about cloud gaming I'm Adam billiard and I'll be sharing my thoughts about where cloud gaming is today and where we all are going with cloud gaming if you are watching this live feel free to ask any questions in the in the chat and if you're watching this in catch up please share your thoughts in the in the comments below so let's kick off with uh you know a simple fact which is that you know cloud is becoming increasingly an important part of our Lives um you know from web applications to to AI to Alexa to Siri um banking login authentication it's a pervasive part of our lives so I guess the question is um why have we not seen uh the full potential of cloud in gaming

so today I want to look at you know why we haven't seen that mass adoption of cloud gaming just yet um so I'm going to be guiding you through a history of cloud gaming um its successes and its failures um and we'll give you a sort of General overview of game streaming from a business and development point of view uh from someone who's who's been in the game technology and streaming um side of the world for for a very long time I want to give a an Insight um into what game uh cloud gaming offers um how it's actually delivered uh and one of the challenges that have stopped uh so many streaming companies from breaking out um so you know mostly I think uh I'm looking to uh arm you to to ask better questions of the vendors as well as evaluate the

potentials uh for uses of cloud gaming you know from a from a business or from a sort of creative point of view um and help you you know get going with um thinking about what a cloud native experience might be so before we get started uh who am I why should you listen to me um well I'm Adam hello um I'm a Brit uh I actually uh spent my High School uh years in Minnesota and that's when I got into programming you know hanging around in Radio Shacks and playing with trs-80s um I've been in the games industry for a very long time uh my first game was published by Atari in 1983. um and of course back in those days uh you not only before you actually wrote the game you had to write the tools to do the art to write the game so it's a very different world um I was co-founder of uh Criterion software uh where we built uh the renderware game engine

um and that was uh for you know for PC Mac PlayStation 2 Playstation 3 uh Xbox Xbox 360 mobile yeah there was a time uh when 50 of all PlayStation 2 titles were using renderware so I worked with amazing teams around the world uh during that time um and when we got uh acquired by uh Electronic Arts um I became the tech CTO and VP in EA and I worked with a lot of the internal teams um and you know amazing teams in EA um more recently I uh co-founded polystream and the idea is that we were building a um a scalable Cloud streaming platform so why care about the cloud you know you may well be sitting there thinking into yourself you know why do I need to know or care uh about what goes on

behind the scenes in in cloud gaming I I just want to write games um sure but I want today to uh lift um show the sort of magic behind the uh the the curtain in some ways um I mean I I still get very excited about you know when you start a stream uh I find it incredible uh that you can start a stream and um you're playing something from the cloud and it's such a great experience and I I know great deal about you know what's going on behind the scenes uh I I still find it absolutely uh magical um but I think there is um clearly you know a huge untapped potential for um game developers with with cloud gaming um and the growth of the market uh isn't slowing down um you know in 5 billion as the revenues you know is the amount that we're going to be passing in 2023 so despite you know Cloud streaming and

cloud gaming so slightly reputation you know I think there's a lot of um reason to be very optimistic about uh the cloud streaming future there is I guess a perception um of uh Cloud streaming somehow defying economics and physics in some in some quarters and and I think um you know for the avoidance of Doubt um it doesn't uh you know power costs money and the speed of light remains constant who knew so do you understand the the sort of uh uh history around cost reach and scale and give insight and knowing uh and and sort of to give insight and also to understand the capabilities the strengths and the limitations of of this medium help will help um I think help me make sort of um informed decisions I guess

so what exactly is Cloud streaming well the cloud in this context is um you know we're referring to an instantly accessible content on demand um so the computer resources you know compute storage uh are available and brought to bear without direct active management from the user and you're playing a game that's running on a server um and it's streaming to your device rather than playing it you know on your um you know your on in your living room on your local device I think the other sort of piece of it is of course that the the nature of the streaming part is that you're playing those apps you know Games movies music where you're not downloading the content first onto your device so you're streaming uh and that whole sort of instant play piece and and just so we're on the same page um I'm talking today about true streaming I'm not going to talk about

um Progressive file downloading schemes I also want to cover some of the moving parts of of cloud streaming um and I think um I think it will Aid understanding and if you've got questions then you know feel free to drop in questions and or email me afterwards so why Cloud streaming is good I think people generally you know today uh understand what the um uh the promise what the advantages of of cloud streaming are you know from instant access to to content to um not needing to manage your own storage not the whole sort of piece of do I delete this game in order to install this new game locally that all goes away with Cloud streaming um you know the ability to to save a game uh from one location and pick it up in a different location potentially on a different device so you could start a game on one device and you could finish the same game uh on a different device

and you know there's things like reduced piracy um there are a couple of other things that we haven't really seen uh to the fore so far which is you know um in in the offerings that we've seen you know to date with things like bringing to Bear huge compute that is in the cloud to give a different experience and also you know the whole sort of uh sort of leveraging the cloud to give never before seen scale experiences you know those two things that we have not seen a great deal to date uh on that front okay GDC diagram time so uh I think if there was um sort of one um diagram one slide uh you you took away from this talk um hopefully you'll take more than just this slide uh this would be the slide I think uh is incredibly uh useful to have insight into what cloud cloud streaming cloud

gaming is so when you're in a uh you know a coffee shop Starbucks um I think we all know that when you're close to the Wi-Fi access point um you get a better experience so if you're close to if you're close to the the access point then typically you know if you're streaming videos or movies they they are less glitchy uh if you download emails it's faster um now the reason that it's the case is that as you go further away from that access point the you know interference there's a whole range of things where uh your data essentially gets dropped and it has to be re-transmitted and that slows everything down and as you come closer to that Wi-Fi access point uh the connection gets better and less packets are dropped and so your your experience gets better so in other words you're closer good far away not so good um the internet of course doesn't work that way particularly um but I think it's a

it's an okay analogy um so in this diagram I'm showing uh on the left hand side we've got uh your your home setup um you've got a router um you may have some Wi-Fi you may have an ISP router provided by the ISP But ultimately you you get that your traffic your game traffic um uh going through a series of routers a series of networks and those hops between routers uh where each of those hops cover a geographic distance um other thing that introduces the the latency um and so you may have you know you may have 5 10 15 20 hops before you get to your actual data center uh where it's going to answer your and send the data back whatever it is um and that's important I mean just sort of roughly to give yourself a sense of it uh crossing the US is kind of roughly 100 milliseconds

um that doesn't suit a tenth of a second it doesn't seem very long but in the context of interactive games where your frame time is 16 milliseconds it's a it's a very big deal um and so that that notion of being close to the data center is very very critical to having a good Cloud experience uh you can't um you know put your data centers thousands of miles away and expect any kind of decent experience it has to be quite close to the users um and to give you a sense of that I think um you know I I ran trace route from my home to uh to Google and just to get to Google's Edge nodes uh for me it was 10 Hops and that's just to get to the edge of um Google's infrastructure and then there'll be a few other hops to you know get to the actual servers themselves

so there are probably two approaches to um Cloud streaming of game streaming um there's sort of pixel streaming uh and command streaming so Hannah's pixel streaming work um well pixel extremely works by sending images using video technology uh the idea is that you read the user input you run the the game loop on the CPU and then you render the the game uh using a GPU that's been provided in the cloud and so that right there means that for every single user you have using your service they need to have a GPU assigned to them so that GPU renders the output as and then when that image is uh complete when the rendering is finished then you take that that image and you encode it using a video encoder which is itself a a process that is quite compute intensive

um and you then stream those video packets to the user and then the reverse happens on the user's device uh those video packets are decoded uh into an image and the process continues uh there's a couple of points one you know that this idea that um you have to wait for the image to finish before you can start encoding it so there's there's some sort of latency that gets introduced uh at that point command streaming is is quite similar um uh so it command streaming takes doesn't send images it sends the data in order to create the image in order to draw the image not the image itself so you read the user input um you run the game Loop and then when you're as you're drawing as your application your game is drawing uh the image those commands used to generate that image are captured and they are compressed and streamed to the user's

device and it's on those on the user's device that um the GPU on the user device is the thing that interprets those that data and does the actual drawing to generate the image so the end result is the same but it's a different way of of um delivering that content I'll go into more depth later on in the slides about command streaming as it sort of pertains to to you know where we're going in the future so you know a sort of a simple summary I suppose um is uh that on the left-hand side you have pixel streaming where you have in the cloud you have a CPU and a GPU for each user um and that GPU uh renders the game and sends it to uh your your computer or your device uh your local GPU which you

have nowadays doesn't really do a lot um because it's just video decoding that's that's happening command streaming is slightly different where you just have a CPU in the cloud you're sending those commands to your your device and it's your local device so your your local devices GPU that does the rendering so you might sort of ask why would you want two different ways of streaming um and I'm going to come back to that but I think in order to understand that I think it's worth sort of going back in time so the announcement of the stadia of stadia at GDC 2019 was the first that you know many had even heard of the concept um but you know cloud cloud streaming and cloud gaming didn't uh appear overnight and there's quite a quite a history and so I think you know understanding uh how

we arrived here in Cloud streaming um and what those choices we made uh both on technology and on the business side um explain more about how we've got to where we are today so I wanted to do a little bit of a timeline and go back and show you know who are the biggest players and what we know and learn from their their technology and business stories foreign right back and when I was doing these slides I mean someone did say to me you know 1980s you know most your engineers weren't born in 1980. um but I think it's kind of an interesting um uh observation that um you know in the 80s you had X11 from MIT you had news from some some Microsystems um and they used command streaming in that they ran the applications in the cloud and then they streamed the commands to to draw the output to your what was called A Thin Client um

X11 was essentially a command protocol news was a bit more groovy uh designed by James Gosling who obviously went on to write a language called Java um it was postscript-based news was postscript-based um in which you sent programs so you sent a program uh to as a bunch of commands to draw what you want to draw so it's it's an interesting kind of piece and I um my postgrad time rather than doing my postcard time I spent an awful lot of my time messing about with PostScript and news so moving on to the 90s uh the big kind of thing I suppose was video on demand VOD um and the idea was you could choose your movies uh to watch and then they were streamed using compressed pixels um you know VHS resolution um there were you know billions of

dollars invested into video technology and uh the delivery of of video on demand you know I remember many times sitting in atsc days meetings thrashing out how VOD and HTML was gonna you know solve the world save the world him um onto the naughties the 2000s um you know I guess the the big piece here that that arrived was h.264 uh h.264 you know it's just an encoder of video but it's it dramatically increased the um the quality and I think you know you probably remember before that you'd be sending these mpeg-2 smudgy little sort of thumbnails that nobody could actually make had a tale of h.264 was a huge step um in in fidelity uh it was also in order to get that Fidelity it increased significantly the um amount of compute you needed to

encode um the video content itself but it's a h264 was a truly stunning piece of engineering um you know from quarter pound filtering to residuals I mean it's what's not to like um so um I think it's the other sort of piece I suppose was at that time was sort of wider adoption of of you know YouTube and Netflix you know started streaming in what 2007 um and computers I mean you know home computers they started you know graphics cards became commodity devices it wasn't whether or not you were doing graphics all PCS all home computers had Graphics chips so 2009 in fact uh on live uh at GDC announced on stage that this would be the last console you would ever buy and the idea of um on live uh was that you would run AAA

games from the cloud uh the natural technology Choice was to take the existing video stream based Solutions and use those to stream not movies but but games interactive games it streamed uh compressed pixels uh to Consumers at HD resolution so 1280 720 um and it also included as you can see in this uh in this picture a a micro console which was the decoder plus it looked after a fedger TV plus it uh um looked after the the controller itself and um you know that was I think just to get a sense of of how radical that was um you have to you know remember that you went into the W bar after that um after that launch and it was wall to wall

um you know self-certified domain experts telling you why it was impossible what they were doing um I know you know I've done the maths it's impossible it doesn't you know it will never work kind of thing and yet you know there are lots of people not not least the engineers on live who had spent their the last few years building this stuff and knew perfectly well it did it works perfectly well so what we learned from online was was you know what were the sort of motivations and sort of expectations I guess that they had and and it was very much around um you know building developers building a more sort of direct relationship with their um you know with their customers with their with their players um and you know the fact that you didn't need to uh have a download or invest in PCS or consoles you know that was the idea it was kind of sort of you know uh make games more accessible in some ways so you know in somebody it was

you were going to purchase games regular games and they were going to be hosted uh with on live uh it was no downloads you could play Anywhere uh you could play on any device um there was some also some added uh kind of interesting social features uh I think you probably remember there was the idea of sharing brag Clips so you could take the last you know 30 seconds of your game uh session and you could share it with your friends uh which is kind of cool and there was Cloud lift which was the sort of feature of engineering that was offered to allow you to play your steam library from your on Live account so this led to uh you know there was a whole history of on live that I think I'm sure many of you know uh so what happened but it led to a a wave of sort of early cloud gaming uh Pioneers in some ways that um were trying to emulate that that

whole um on live on live piece so you know there's a selection I've got uh here of um by no means exhaustive I think the interesting thing is that you know when you look at uh the the history of of cloud gaming you know many um you'll notice of either close their doors pivoted or been absorbed into bigger companies and you hear very little uh more about them so what happened I think the the to fully understand you know what happened uh during what I would call uh you know the rise and stall of cloud gaming uh you need to look quite carefully at what the pros and cons of uh Cloud streaming today are but before I do that um there's one entrance uh that I

haven't mentioned uh that uh should have been should have differed from all the others because it wasn't a startup um stadia so stadia announced in GDC 2019 that's a decade after on live uh launched at GDC and it promised you know super powerful data centers delivering Blockbuster Games you know no need for downloads uh any device would be supported uh and it was particularly again aimed at people who didn't have access to high-end PCS and consoles does that sound familiar you know in many ways uh stadia was a carbon copy of the online Playbook um but I was personally you know super optimistic that uh when it launched that you know Google did bring something quite different and new to the table I mean first off was Will beating you know

infrastructure and deployment uh expertise um there was the the promise of you know tight integration with with existing services like like YouTube very ambitious aspirations they they talked about about delivering 4K and 8K resolutions and lastly and probably most importantly they brought an infinite money tree to pay for it all so given all that you know why is it not working well I think um it I think it's an interesting kind of question as to why have most cloud gaming stream offerings uh turn to pixel streaming um more than anything else well the pros of pixel streaming you know it's a known technology from from the introduction of uh VOD via video on demand uh it's pretty easy to onboard content um reduced sort of upfront engineering

you know consumers Have and Have video players or things like video players already uh it plays on a potato and that that is an interesting piece I mean so the way video technology works it's designed from the ground up uh to make it so that while the encoding of content may be fiercely expensive from a compute point of view the decode of that content is very very low compute requirements so it was very much this idea that that you could run um you know fantastic content on super super low end you know 10 year old um laptops and plus you know existing Network transport firewall videos and a known quantity I suppose okay so let's just look at the the you know I mean because given all that you you kind of Wonder well that sounds great the cons though are kind of interesting and and perhaps are more

relevant to to you know the history of of cloud gaming you can't use cdns and caching when you're doing pixel streaming and the reason is is because it's interactive you don't know what the next frame is going to be um you can't sort of uh cache uh 10 seconds in the future because you don't know what the what the player is going to do so not being able to use cdns and uh caching you know which are in some ways you know the very heart of what the internet is built on is is a is a bit of a blow in some ways um you know it requires also you know ongoing GPU upgrades so if your if your model is you can turn your game up to ultra and you can run it you know insane settings and I'm I'm going to stream to a potato uh that's great um But ultimately uh it does mean that every few years or everyone not including Google are

going to have to replace all those gpus and get new gpus um so just to stay current so you get a new game the you know the the next AAA game that's using all the latest you know Cutting Edge GPU features and if your if your model is you can stream that at Ultra it does mean that you're going to have to replace all those you know many many thousands of gpus the other sort of piece I suppose is you know as we move towards uh increasing resolution the egress the amount of bandwidth you need for an 8K display is significantly more than a 4K display um you know that's a piece that that's it's not a Now problem but I think you know you can see coming down the road that that is going to be quite a challenge for for Pixel streaming uh it's worth talking about time of day and the time of day is an interesting one um so the difficulty with um with interactive content is as I said earlier

you need to be close to your your users uh you need to be physically close to them you can't for example um from Midnight to 5 a.m in the morning use your gpus your expensive gpus you bought in San Jose to serve the East Coast or your European customers you have to have gpus in all locations in order to serve your your um your customers and that's problematic um you know pixel streaming in some ways is a little bit like a gym subscription model which is you kind of want lots and lots of Subs uh you just really don't want them all turning up on January the first for the for the jogging machines you know because uh it's you can have a capacity problem uh if a lot of people turn up and all want to play games at the same time I think it's uh kind of telling in some ways that the real-time conference last month uh you know a panel with a lot of luminaries um you know significant uh players in the get sort of cloud gaming

um and Cloud streaming world uh you know said categorically that that pixel streaming is is not and probably will never be a b2c offering just because you can't predict user numbers you can't scale on demand so you know the fact that each user needs a GPU in the cloud uh is makes it so it's difficult to scale okay the challenges of pixel streaming um is comes down to power and scale so for power and I mean electricity um for you know Cooling and running these things uh yeah the average cost of running a GPU stream is going to be of the order of 50 cents per hour per user some people do you know are you know more efficient some are less but it's of that sort of order and if you just pause for a moment and think uh as a business uh if you are going to cost you 50 cents

per hour per user to deliver content that that's quite a lot of modern monetization that you're going to need to do um to make that a viable business so you know notwithstanding Google's uh quantum entanglement uh work they talked about at the stadium 2019 launch um you know it's going to be over the order of that price the other piece of this is uh scale so you know if you look at the the launch window for a triple a title um you know today it's in the multiples of millions um so you know you go and look at steam stats and see you know how many millions of people are turning up to playing some of these new games you contrast that with the CCU capacity you know the the how many people can simultaneously play um on some of the pixel streaming vendors and it's in multiples of a thousand of thousands

uh they're just it's a huge um expectation mismatch between um the the consumer millions of consumers playing games and the quite limited scale that um gpus are able to to scale to so you say wait what about Netflix they serve millions of customers well Netflix is is uh serving movie files essentially um and that is not the same as streaming interactive uh content it's just not the same thing uh you know for starters you can if you've got a movie you can take weeks um permuting all combinations of your encoder um settings so that you can get the most optimal uh compression for this particular movie um because you you have the whole movie and you can run your encoder let's say

four weeks to get that thing and it's all done offline once you've done it you're saving that file you're done uh in cloud gaming you know you have of the order of eight milliseconds in which to make those choices it's a very very different um kettle of fish uh the other thing with Netflix of course is that you can use cdns and caches and so you can cache portions segments of your movie around the country around the world so that they're close to users um and and that is something that again is not available to uh interactive cloud gaming because of the nature of uh the interactivity you can't predict you know what what the next uh image is going to be and lastly um you know movie encoders have a notion of B frames B frames essentially reference um think of it as you know a cache warming for macro blocks uh it's a it's a frame uh in the in the video stream

that looks forward uh to um video information that's not yet shown but will be shown in a short period of time for movies you have that information because you have the whole movie for a game you clearly don't have the very next frame even so B frames are out um and and that has a pretty significant effect on both quality and and bandwidth Netflix you know can as I say they can take weeks and weeks um compressing these things to save egress and uh it's a very different thing I think the other sort of piece is with with Netflix is uh if your movie starts 300 milliseconds late it's not really a problem um you know it's it's just not a significant uh problem where uh you've got something that starts a few hundred milliseconds late you probably wouldn't even notice and the last thing I think is is that you know Dash players um you know the the movie players if you like um they themselves will cache seconds of content to completely hide any network Jitter that you may have so Netflix very

different proposition um if you know colleagues boss says what about Netflix just say no it's very very different so what do we need to do to succeed well I want to focus now on on the future in some ways and and it's a very positive future uh you know I want to remind just sort of a reminder I suppose is what we're talking about here is commodity um cloud gaming so they've audience of doubt you know there are plenty of Enterprise uh uses of pixel streaming um and they're quite happy to pay you know five dollars per hour per user for a timed experience that that's not what we're talking about here we're here we're talking about something which is an on-demand commodity for millions of consumers so there are two sort of key issues that need to be addressed one is delivery and two is content itself

um on the delivery side you know the three sort of pillars of this are cost reach scale so on a cost basis if you want to have a commodity consumer facing service it has to be pennies per hour per user you cannot build a um viable business if uh and if you're paying 50 cents per user per hour it needs to be pennies it needs to be a significantly small amount of money secondly there's reach there's the whole idea of being you as I talked about earlier you need to be close to your your players your users your consumers um in order to make this happen and that means that being able to spin up um capacity close by um when uh as necessary and I think you know that that um sort of scale piece of you know even today you look at you know um a stadium it's full of of Spectators for example

and then it's empty and that's sort of the landscape I I think we're going to be uh looking at for for cloud gaming um that's it's kind of spiky um and it has to last a Euro and I think we've seen that already I think with some of the digital uh events that we've seen um but by the nature they are they are transient so um the other sort of aspect of this is content um taking existing content and and simply sort of uh putting it into the cloud take a game you put it in the cloud that's just not enough you know reducing friction through uh instant play um you know and we all get frustrated when we sit down and you know you want to play a game and it says I'm gonna do patches for the next 40 minutes yeah it's irritating um but ultimately you know the friction that comes from instant play you know it's a great feature it's just not enough it's not compelling enough and

the last thing I suppose is is you know offering the same old business models that we used for for physical Goods as much as anything it hasn't really changed a great deal uh and using those same old business model is just it's just not enough so we need to uh play to the strengths of the medium so what are those strengths well um I think we have many kind of existence proofs uh that synchronous social has been very well served by by Cloud infrastructure um so be it you know Facebook twitch uh Discord you know they've all uh they are cloud-based services and they work very very well deals so you know the other sort of piece I think of of uh the strength of this medium of cloud is composable compute and the idea of being able to uh scale up and down dynamically to um support different demands it's something that's at the very core of

elastic compute and it's something we have to leverage um not from a necessarily partly from a capacity point of view but also from the kind of experiences that we want to to to build and I think the other sort of I've said solved Network netcode um you know one thing that's interesting is once your game client is running in a data center uh the the interconnect between and the sort of Communication channel between your game clients and the other players for example can be super low latency and you know super high performance I mean data center uh the intra data center sort of connectivity you know it starts at the bottom as very very high and then goes up from there so leveraging the fact that you're running these uh these experiences these games inside the data center so that you can make use of of all those cloud services but at super low latency is something I think is

really so does play to the strength of of the the um of the medium so you know it's about building Cloud native experiences and and it's already happening to today to a degree so if you look back on um absolutely 2020 hindsight what have we learned from from you know cloud gaming to date uh I think you know certainly the early straight line of cloud gaming which was throw away your console and trying to reproduce that console experience in the cloud uh now looks um you know a little bit social sizes I think you know and it and it's sort of essentially trying to cannibalize a different skew by putting it in the cloud instead of trying to create experiences which which uh uh play to the strengths of uh Cloud uh rather than just try and do what a

console does I think uh as far as the dues are concerned rather than the don'ts um you know thinking of uh beyond the traditional sort of um game mechanic win mechanic that we've had with us for you know 30 plus years and just to sort of create unique experiences and Thinking Beyond those traditional audiences so it's about not necessarily um uh building experiences for gamers but building it for non-gamers and so you know we have these things and we see examples of it you know today of having game worlds which you're exploring in-game content for entertainment um you know and it's uh creating a a cohort of people that are um Spectators and they are sort of orthogonal to the player mechanic they they're completely sort of orthogonal to the gameplay mechanic you can have great player mechanics and you can have spectator mechanics and and they're they're not necessarily the same thing at all the difference is The Spectator

mechanic is you know hundreds of times larger than your player mechanic I think the other sort of piece is is leveraging uh as a do is leveraging Cloud compute and AI so again as I said before that you know when you're running your game inside the data center that means you have available to you a whole bunch of services and bringing to Bear compute that is super low latency and can you know be it from you know deeper behaviors to you know more satisfying kind of cinematic experiences to you know even faster content creation of using Ai and there's a bunch of stuff out there already that does that but what changes I think is when you're running um the the game itself in the data center so you're not doing that back all the way back from uh from the homes um so you know the two things I think we need uh for a future of cloud gaming is one you need it to be business viable

um and secondly it needs to scale with demand so before we we look at you know where we might go with cloud gaming I think it's worth to revisit how we might deliver to Consumers so just this is a techie slide it's just a slide we'll we'll get back to uh um more broad stuff afterwards um but I think it's interesting to look at command streaming because I think when I was talking earlier I was talking about how uh command streaming um works and I just want to take a little little small step deeper into um the tech because I think uh it helps unlock uh a viable business and scalable elements uh you know that that we need so the way command streaming works is uh you choose your streams visual Fidelity based on each user's device GPU in other

words um if the visual Fidelity you get is based on what GPU you have um and so that different people with different gpus get different uh can have different experiences it can utilize uh caches and cdns which is obviously good for reducing egress uh command streaming workloads are containerized which means you can dynamically scale they're also kind of obviously Cloud agnostic because you're not chasing the availability of gpus you're chasing the availability of CPUs and there are you know hundreds of millions of those and so you can get close to the user and again back to the uh earlier is um excuse me um is being close to the user reducing that distance is an important part of giving a great environment a great experience so less hops good um you know the transport uses a sort of bbr style you know it's a

round-trip time uh congestion control but it interestingly also accommodates having sort of Burst Mode uh chunking a bit like you know Dash players because of the nature of of command streaming not being pixel based and I think kind of most significantly of all uh is the cost to serve the CTS for command streaming is in some sense you can see it as being shared between the data center and the user device so what does it all mean well um command streaming I think can enable creations at massive concurrent uh engagements and they can be built from the ground up I mean so we know that uh a little bit about the technology behind it um there's plenty of existence proofs of command streaming delivering content and there are absolutely examples of of existing content running poorly um so we need to be designing uh these

experiences uh from the ground up to you know play to the strengths of of cloud look at those new spectator mechanics that are Beyond uh traditional Gamers um how do we bring in uh different people to enjoy that experience and playing to the strengths of command streaming which is the way in which I think you we can deliver at scale so that I think leads to a lot of benefits and you know as much as anything it's it's low friction access to um uh Synchro social experiences it's bringing non-gamers as I say the business models to monetize non-players is something that I think is a really exciting and and uh something very different from the the slightly sort of Sledgehammer thing we've done up to now um and it's moving beyond uh games and the sort of wind mechanic we tended to do for for our our games so the challenge is uh are that you know

moving from experience to experience is going to require a whole bunch of interop between content vendors uh you know and that means things like you know shared industry standards what do we do about walled Gardens you know the reality is that a lot of wall Garden experiences that are great out there you know how do they get incorporated into something where I can move seamlessly from one experience to another you know and also I think it's about um supporting developers about how they create for cloud and uh command streaming there's also I think the challenge is that there's a sort of reluctance to move away from expensive pixel streaming infrastructures because they're because they're familiar so if we look today um you know there are quite a few games out there that have spectator modes um you know the next step will be having thousands of Spectators viewing uh viewing the action and um using Cloud streaming so again you've

got a a player mechanic running an orthogonal uh spectator mechanic that's running completely orthogonal to that but in their thousands you know what does that mean um spec cases become a cohort you know as I say um and they have their own play mechanic uh it's about meeting a virtual virtual spaces it's about moving seamlessly from one uh experience to another and I think you know it's if you look at the trend right now it isn't uh Strictly Come on streaming it's more sort of metaverse style that I've tried what do we need to do to make it so that we can move uh between these experiences um it without um you know and do it do it at a huge scale so I think there's there's a whole bunch of interesting companies uh that are doing great stuff um you know there's you know unreal hadian improbable I mean it's it's interesting that that game engines in some way is a misnomer and increasingly

it's a a collection of tools that for for powering you know interactive experiences um yeah High Fidelity I think is is an interesting one I'm sort of calling out in some ways because it's it's they're looking at the audio and and how do you deliver uh spatialized binaural uh audio to massive 3d um 3D audiences and experiences um the typical guys uh you know again the whole client to Cloud Revolution work they're doing and I urge you to go and look at some of the stuff they're doing about how do you in practice you know solve some of these interrupt problems across a wide range of Industries um and platforms and I'm excited by you know Mainframe Industries um you know the things like scavenger and uh from uh midwinter and entertainment avakin life Manticore Roblox I mean all these guys are doing incredible stuff with with regard to um you know bringing in a uh a whole

bunch of new participants that weren't there before um I think there is a a piece here though about you know you can't join these things unless you've got uh the the um the wall Garden installed and you know we've all read that um well please ensure that you download you know the the game uh well in advance of the concert I think is the is the phrase uh but we can stitch these together uh as a sort of heterogeneous metaverse experience I think you know with things like cloud gaming so my key gate um so takeaways I suppose is you know content needs to be Cloud native uh we need to play to the strengths of cloud compute we need to design for millions of consumers and Beyond the traditional game game mechanic win mechanic uh and even the gamers that we've we've traditionally targeted uh and I think there's a huge opportunity for transactions between participants yeah

what does that actually mean so content delivery needs to be uh to scale to sort of consumer level numbers in order to kind of deliver uh these kind of experiences so thank you very much um you know if you have questions uh you want to chat about stuff feel free to drop me a line uh you can find me on LinkedIn thank you [Music]