AWS re:Invent 2020: Live content production on the AWS Cloud

AWS re:Invent 2020: Live content production on the AWS Cloud

Show Video

hello and welcome this session is mds 204 we're talking about live content production on the aws cloud my name is claire saudi i'm a specialist solutions architect with amazon web services and i'm joined today by mike crong the vice president of advanced technology from grass valley and also claudia souza the director of solutions architects from grass valley grass valley is an aws solutions partner and a provider of advanced technologies for broadcast and media and entertainment use cases covering the entire media supply chain from cameras and acquisition systems all the way through to consumer distribution this session assumes some basic knowledge of media and entertainment we assume that you have some understanding of the basic functions involved in a media supply chain and we assume we have a you have a basic understanding of the mechanisms for video manipulation video compression and video transport the agenda for today's session we'll be talking about what is live video production and how does it work but more importantly what are some performance technologies for live video production what are the essences of what a technology stack needs to be able to do to deliver live video production on the cloud we're going to look at a customer use case with electronic arts and how electronic arts has gone about delivering the reality of live production on aws and then my colleagues from grass valley are going to take you through the aws architecture they've got in place to deliver a successful live production using grass valley amp for electronic arts then at the end of the session we're quickly going to talk through some of the underlying aws technologies that are available for you to use as a developer or a solutions architect if you are building live production solutions on the aws cloud so what is live video production well production refers to the process of composing together all the different interstitial elements that together make up a production and that includes all the different camera angles all the different microphones all the stills and graphics all the titles and all the chroma keying effects all in real time as the content is being delivered to the consumers because the consumers are seeing the action at very much the same time as you're producing it as a content producer that means you only get one go at this this is happening live on the flyer you're creating a live video as it's been delivered to your consumers and there's no opportunity to go back and edit it afterwards this is historically a part of the content supply chain which has been dominated by on-premises hardware and part of the reason for that is that the bit rates are very high and demands a very low level of latency in terms of video processing but there are constraints associated that with that we know that hardware-based video production is capex intensive we know that it has poor return on assets and part of that is because the workloads are often very bursty live production often occurs when live action is occurring when there's a perhaps an election in progress or a sporting event that's of significance or a cultural event that is of significance to to audiences that's when the live production occurs but when none of those things are happening as a as an owner of assets you've got assets sitting out or doing nothing i mean you're not able to monetize those assets effectively so this aligns very well with aws cloud and the use of cloud technologies for delivering better economic value but also delivering more options of content a more diverse range of content to customers through opex-based workflows this is perhaps the last challenge in the media and entertainment supply chain we know that there's a fairly high degree of maturity already with ott and video distribution and that reflects maturity of internet technologies and cdn technologies and video encoding apparatuses lydian playout is relatively mature at this point in time we're probably past the stage of early adopters early adopters taking on linear play out in the cloud but live production is still perhaps the final frontier and the biggest challenge the ability to switch between live cameras on the fly or perhaps if you're doing an esports or gaming tournament the ability to cover multiple gamers at one time and add commentary to that and add all your visual effects at the same time at the same time you deliver that to audiences it is a challenge and this is one where aws is continuing to grow into this space with our customers today so what do performance technologies for live production look like on aws well they need to be highly reliable we know that a customer who has purchased rights to a high-profile sports event such as an olympic games or a super bowl has invested a significant amount of money and they've only got a small amount of time to to deliver results to their audiences and so reliability becomes critical seconds matter in broadcast live production also needs to be cost efficient moving to cloud does it shouldn't encumber a customer further from what their traditional broadcast operations look like and it needs to be low latency we have operators who are responding in real time to what is happening on the screen they're providing cues to camera operators they're providing cues to talent when they should begin speaking they're providing production cues in real time so latency becomes critical at the same time the reason for this journey is flexibility and scalability uh customers looking for that ability to spin up more control rooms on demand and spin up more production capability when they're busy when productions are in flight and scale them down and shed those resources when production's not in flight so the goal there is to get flexible scalable infrastructure without a compromise on reliability cost efficiency or latency so let's take a look at one customer who's working on aws today doing live production how grass valley have worked with electronic arts to make this a reality thanks claire so electronic arts is a global leader in digital interactive entertainment ea develops and delivers games content and online services to more than 300 million registered players around the world part of that experience they provide for their players and fans is broadcasting tournaments around the globe for their many properties including madden fifa and apex legends ea has an entire team dedicated to producing and broadcasting all their tournaments the competitive gaming entertainment team had begun leveraging cloud-based production tools earlier this year but they were looking to improve their experience producing the tournaments as well as the end result that they were delivering to their fans so let's look at some of the requirements that they had that they were going out and talking to different vendors to see what they needed uh done so basically they have obviously a very distributed workforce so they have creatives as well as operational staff that are all over the globe so they needed a way to manage that distributed workforce they also wanted to make sure that the product that they delivered would have the broadcast quality similar to what they would do in a traditional broadcast environment even though they were doing this via the cloud most importantly because they have all these different events going on globally so you can have some events in india or aipac or latam or even here in north america all happening simultaneously so they needed the ability to pivot in scale to meet the demands that were growing and shrinking around the world in a way where they didn't have to ship equipment or people anywhere but still got that broadcast quality so to enable that it was important that they found a way to have extensive probing and monitoring capabilities they wanted to ensure that even though the production was being done in this new way they would be able to keep an eye on everything that was going on in many different productions simultaneously and address any issues as they happened so they took a very uh solutions based approach to how to solve this need that they had they spent months and months researching and looking into different solutions and as a matter of fact since they have different games and different properties they were each using a different type of remote production cloud-based and some more traditional remote production but they really wanted to see how they all stacked up against each other they were also keeping an eye out to what others were doing in the industry and that's how they came to find grass valley and our amp agile media processing platform product they heard of others using it in their industry they came to us to investigate whether our technology would significantly improve either the quality of what they were delivering or the process by which they were delivering it and obviously by this uh this gate this thing today that we're talking about they did choose the agile media processing platform so i'm going to talk a little bit about why what were the key factors they were looking at and what was important to them in making that choice so as they started to investigate and eventually even do a poc with us what they were really looking at and what impressed them was the ability to have operator specific user interfaces they wanted to make sure that each of the different people working on the production could have their own user interface and really focus on their key part of what they were contributing to the production without having to see what everybody else was doing yet have it all in a unified universal user interface so that they could change what they were doing to different events all with that same repeatable interface they also needed that low latency high quality monitoring i talked about earlier and that was a key aspect of what they needed us to be able to do for them obviously when you are doing these uh geographically diverse contributors and events it's really important to have an ability to manage the delays that are inherent in any cloud-based live production and that's where intelligent timing management came in we enabled them to be able to have a professional level broadcast production all live all in the cloud because we're managing how all these different assets are coming in and making sure that the timing and the latency is not causing an issue for them also important was having a wide enough breadth of tools they wanted to make sure that they didn't have to compromise their production by just using what was available or hobbling together different things from different people so they were looking for somebody who had enough of their production chain that they could really um work in a way that they had become accustomed to in their traditional productions and we have enough of the chain everything from the um different tool sets to the different operator interfaces that they were used to so now let's talk a little bit about the scope of their production what were they really looking at and what was their typical production that they needed us to assist them with so right now what they what apex legend specifically was looking at is that they have eight virtual cameras and these are their observer stations they also have eight contribution links from all of their talent that are contributing to the show and they need to be able to do all the live switching graphics audio and replay regardless of where these contributions are coming from or the operators are around the world and of course they wanted all this to be a hundred percent cloud-based processing so i'm going to hand it over now to our vice president of advanced technology mike ronk so he can dive deep into some more of the technical details behind how all this is done with our agile media processing platform so mike take it away thanks claudia so let's take a look at the solution that ea apex selected it's based on grass valley's agile media processing platform or amp and it's the first cloud-based sas platform capable of addressing broadcast grade live production applications so what do we mean by that well like a a truly elastic cloud system it's agile it's elastic it's very flexible uh it's scalable at web scale uh and we're able to release new apps and services continuously but it also meets the requirements that claire outlined in in the beginning for live production and that's how we why we call it broadcast grade so how did electronic arts apex use this uh as as their solution what i'm showing you here is a system footprint it's pretty hard to read there's a 169 separate microservice based applications working in concert together to pull off their live production but let me just diagram it a little bit for you so first of all on the left in the blue you see the application services that are running on computers on the ground for for contribution into the cloud and then on the right are our five ec2 instances on which a whole bunch of applications amp applications are are running and so on the ground uh they have talent in in the sport esports world they call those casters and they have observers these are essentially uh the cameras uh they're virtual so that people are flying around in the game in 3d space and unlike the real players in apex they're just not shooting anybody and so those those camera feeds and the headshots with cam with with microphones from the talent are all being fed into the cloud at contribution quality typically 15 to 20 megabits per second and then in the cloud the first thing that happens is all those observer feeds those cameras are sub switched to produce the kind of a raw feed of the program cut which is also fed back then to the casters in low latency monitoring so they can see the game and comment on it from wherever they are and then from that sub switch there's there's other production values so creating effects uh multi boxes with graphic backgrounds uh are done in one instance uh and then there's also a main pcr that's kind of stitching all this together with with graphics to a final program out but from there in a broadcast typically there's a distribution stage network operations uh sometimes historically called master control where specific branding for each distribution channel so in electronics arts case they distribute via twitch uh and youtube and those are separate channels and there's other channels that they do and there's an additional ec2 that's taking in signals that they want to use in the broadcast and feeding things out uh to other people that that need to see them so this really formulates the entirety of of their solution that they can spin up and down uh at will and connect to really people's homes all over the world uh and to produce a coherent uh video uh broadcast so what's the architecture uh behind this how does this done on on aws well amp has two fundamental architectural planes a management or orchestration layer and then a data plane and the management plane is hosted across three availability zones for resiliency and availability leveraging kubernetes and uses native aws services such as elastic search service and alaska cache and in fact some of the services that we run in the in the ec2 containers that you see in the management plane there uh we are moving to uh aws native services for additional uh scalability as as the customer demand for for this is very very high um also on then on the data plane these are the larger graphics instances ec2 instances that run the video processing and do the the computational lifting for for video and and audio and these communicate back into the orchestration layer and the customer literally can deploy those anywhere they want uh in any aws data center uh across across the world and in fact uh that gives them the the geographic pivotability that uh electronics arts uh was seeking that claudia was speaking about so that's the the architecture um and so the results of of this solution for ea sports well they had a quick learning curve they started the poc about five weeks before their first on-air training everybody a new system building it up from scratch uh really phenomenal versus what has to typically happen um in in a more bricks and mortar type type of situation with traditional equipment they improved the production quality of their broadcast they were able to automate further using apis to turn things on and off more more easily and smoothly the one of the feedbacks was we had great great support and more of a devops model um so we're constantly being able to interact with them and work with them through that process and it's really fostered an ongoing collaboration which i think is indicative in this early stage for live production in the cloud with both grass valley and aws and working working with customers is a very dynamic and collaborative environment and so to give you a little bit more flavor of what it's like to do live production in the cloud we have a short video here that i'd like to share with you this is from a internal grass valley video production so we're really eating our own dog food with amp internally as we as we do communications and ian fletcher who's critical in the development of app is our chief applications designer will take you through a little tour of what it's like to produce on amp [Music] [Applause] [Music] coffee well we've just finished our gb live engage 2020 event which was produced entirely on amp so all of this was done using the public internet so these dotted lines represent the the feeds going from the on-premise compute nodes which are capturing the video across the public internet to get to the aws data center and in this situation because these are contribution quality feeds we're using a low latency and codec at 15 megabits in this case and we also have some wrist dialed in for a bit of stream protection which added a little bit of latency but it's worth it to guarantee you're going to get those solid pictures that you saw but one of the things you might have noticed is that there's very little delay when you see the presenter communicating to basically someone who's halfway around the world and that's one of the powers of amp is the way it manages those latencies and you can't cheat time but you can manage it the amp ecosystem will automatically time align those for us so that we're able to um basically make sure everything everybody talks nobody talks across each other and they can also be pretty interactive without getting any of those sort of embarrassing pauses that you often get with satellite contribution with traditional broadcast systems so over here in the uk i was kind of handling master control i was handling audio using another one of our user interfaces for managing audio levels i could watch the multi-viewer using one of our low latency flow monitors which is also a key part of the amp solution as could the director and the td in florida they'd also got this low latency return which they put up on there on their monitor wall so they could see what was going on we also made extensive use of our gvui control surfaces which let us bring lots of different components onto the same screen so in this case i had the main production switcher larissa over in um in belgium she was doing all the clip playback and we used quite a lot of our clip players because we had not only a b and c for playing in the main sort of pre-recorded elements of the show but we also had graphics backgrounds we were rolling we had name supers so there was quite a lot quite a lot going on so we had six clip players in total so how did what were the benefits that ea apex realized using the the technology that that ian was describing there were first of all they really nailed the whole geographic diversity uh goal they're able to spin up uh the production wherever they need have uh talent and operators uh country and and the game players contributing from wherever they are really in in in the world uh it does give them that that real geographic diversity that they were seeking because their audience uh is worldwide their players are worldwide it is truly elastic uh it's on a pay-as-you-go model uh when they're not don't have a a game they're not using it uh and and it's you know buy twice a month type of thing so it's a perfect elastic use case um for electronic arts uh it also is enabling not just remote production but work from home where literally every single operator every single talent that's contributing is working from their home separated but they can do a tight-knit show uh with great communication and with the managed latency make it look really really professional uh without any real flubs and so uh the other benefit is is the tools that are provided are intuitive from somebody that's used to doing professional broadcast and that helps it's a great benefit and was contributing to the quick learning curve so that's the solution uh that ea apex selected it's been a great experience working with them and working with aws on on this and other projects and we're very excited about the potential for live production in the cloud so with that i'll turn it back to you claire thanks for sharing that mike so that's an example of how one customer has gone about delivering the reality of live video production on aws let's talk some more about some of the technologies that are available to you as a video developer as you're building your own applications on aws so one of the biggest challenges with live video production is performance it's delivering the amount of throughput required we know that potentially a high definition signal can consume up to 12 gigabits per second traveling from one instance in the 80 cloud to another instance in the aws cloud we know that live production customers demand uncompressed uh performance and uncompressed picture quality and we know that they expect quite often near zero latency to achieve that and we know that there's expectations around high reliability so aws cloud digital interface is a network fabric that's available to you as a developer if you are building uh content applications it leverages the elastic fabric adapter which is our high performance compute network interface card and it delivers video from instance to instance to instance reliably with low latency and that allows you to build more complex applications and more complex architectures and configurations that provide interoperability between perhaps one application stack and another application stack as you might be integrating together for example audio mixing with video switching with interstitial playout and maybe graphics engines all those components can be stitched together using aws cloud digital interface this is a software development kit which is available to you today it's available on the aws github and there's download links on the aws website it's usable as both a linux sdk or a windows sdk and it's also available open source and what that means is you can take the upper part of the sdk and use it in a hybrid configuration if you have some of your video operations running on premises and some of your video operations you're looking to run in aws you can build hybrid configurations because the sdk is openly openly sourced it does allow you to build interoperable fabrics and adoption hopefully will continue to increase increase over the coming months so that you can you can work with other vendors if you are building a video application and provide interoperability so today for this session we've talked a little bit about the importance of the aws cloud and how that will interact in the future and today with live production we've talked a little bit about one customer use case with electronic arts and how they've made a reality of live production aws cloud with the support of grass valley and we've talked about some of the technologies you can use as a developer as you're building video applications in the cloud if you want to drill down a little bit deeper we're sorry we can't be with you in person today to talk through your use case for live video production the cloud after this session but we do have our email addresses available and you can reach out to us at any time to talk about live video production or of course if you have an aws account manager you can chat with them as well so this session has been mds-204 live video production on the cloud thank you for watching

2021-02-13 03:28

Show Video

Other news