[Streamroot] Whitepaper peer assisted adaptive streaming
•
2 likes•493 views
This white paper introduces a new peer-assisted approach to video streaming designed to overcome limitations of content delivery networks (CDNs). It explains how peer-to-peer streaming can represent a key advantage for broadcasters by enabling them to scale up, improve quality, and handle traffic peaks while reducing costs and network burden. A case study showed the solution achieved up to 58% peer streaming and ensured continued streaming for 50% of users during a server outage. Peer-assisted streaming optimizes video delivery as demand increases by leveraging growing numbers of viewers to share content.
Report
Share
[Streamroot] Whitepaper peer assisted adaptive streaming
- 1. PEER-ASSISTED ADAPTATIVE STREAMING: THE KEY TO MANAGING EVER-GROWING ONLINE VIDEO TRAFFIC W H I T E P A P E R
- 2. FASTER DELIVERY + LESS BANDWIDTH + INSTANT SCALING INTENDED READERS This White Paper is intended for online broadcasters, CDNs, video platforms and professionals all along the online video distribution chain.
- 3. 3 CONTENTS CONTEXT: A BOOMING ONLINE VIDEO MARKET IN SEARCH OF TOMORROW’S SOLUTIONS PEER-ASSISTED DELIVERY: OPTIMIZE VIDEO PLAYOUT AND MINIMIZE INFRASTRUCTURE RISKS CASE STUDIES: OVER 50% PEER-TO-PEER STREAMING ON A LARGE VOD PLATFORM FINAL REMARKS 1 2 3 4
- 4. 4 ABSTRACT This White Paper introduces a new peer-based approach to video streaming designed to overcome the limitations imposed by current Content Delivery Network (CDN) infrastructures and distribution techniques. It explains through statistics and case studies how peer- assisted video delivery can represent a key competitive advantage, enabling broadcasters to scale up, improve quality of streams and handle traffic peaks, all while reducing costs and lightening the burden on saturated network infrastructures.
- 5. 5 CONTEXT: A BOOMING ONLINE VIDEO MARKET IN SEARCH OF TOMORROW’S SOLUTIONS A GROWING ONLINE VIDEO MARKET The online video market is experiencing exponential growth with the combined effect of skyrocketing consumption and increasing video quality. Internet traffic is poised to more than double in the next five years. Today video content represents over 60% of worldwide consumer traffic. With the growing use of mobile devices and tablets, this figure is set to jump to up to 90% by 2018.i According to a recent Ericsson report, video traffic will grow faster than any other mobile segment – 13-fold by 2019 – and will soon represent over half of mobile data use.ii 1 At the same time, video files themselves are growing. In a market where performance is essential, 4K, or ultra-high definition (UHD), is quickly becoming the benchmark.iii As consumers come to expect heightened quality, video distributors will require more and more bandwidth, as well as infrastructures capable of receiving larger files and increasing amounts of traffic, if they hope to ensure a satisfactory end-user experience. Figure 2: Projected growth in mobile video and total mobile trafficFigure 1: Projected growth in video and total Internet traffic i. Cisco® 2014 Visual Networking Index ii. Jim O’Neill, “Mobile video traffic to grow 13X by 2019, straining already congested networks” 09 June 2014. http://www.ooyala.com/ videomind/ blog/mobile-video-traffic-grow-13x- 2019-straining-already-congested- networks#sthash.zRB5X5f0.dpuf iii. A UHD file is 16 times larger than the same file in HD.
- 6. 6 A UNICAST MODEL SHOWING ITS LIMITS Current distribution models are already starting to show their limits. As Akamai’s Kurt Michel observes, “Increasing broadband penetration and faster devices make high-quality viewing experiences possible, but network latency issues and heavy traffic loads can often result in disappointing video streaming performance, with frequent pauses for rebuffering.”iv Even more troubling, outages are multiplying as platforms struggle to cope with unprecedented demand. Even the sector’s biggest and most lucrative actors – those that invest huge sums in top-tier CDNs to ensure content is delivered to millions of users – remain highly vulnerable. Look no further than the last few months: Just weeks after problems with the True Detective finale, HBO Go suffered a crippling outage potentially affecting millions of viewers during the Game of Thrones season 4 premiere; ESPN crashed when 1.4 million simultaneous viewers attempted to watch the US-Germany World Cup match; ABC’s debut live stream of the Oscars was down for most of the evening.v Traditional unicast protocols based on one-to-one relations between client and server have proven insufficient, with frequent network over-capacity and congestion. Content delivery networks are racing to palliate the situation by multiplying peering points and surrogate servers around the globe. This solution, however, is not scalable ad infinitum and risks to ultimately fail to provide the reliability sought at an acceptable cost. What’s more, while content delivery networks can take much of the burden off of a provider’s origin infrastructure and help deliver content to users faster, CDN infrastructure is shared among multiple customers. If one client needs to broadcast a huge sporting or political event, will other customers be forced to suffer? The bandwidth wars have only just begun, and those with the architecture will decide which videos will be delivered at what speed (see the Verizon / Netflix conflict). As telcos vie to compete with traditional CDNs, more and more companies are looking for a slice of the pie, and bandwidth-heavy video content providers will be at the mercy of potentially unfavorable pricing and distribution policies. iv. https://blogs.akamai.com/2013/01/ live-video-streaming-that-can-handle- traffic-spikes-the-challenge.html v. For more information on these exa- mples, see: http://techcrunch.com/2014/04/06/ hbo-go-still-recovering-from-outage-du- ring-game-of-thrones-premiere/ http://variety.com/2014/digital/news/ espn-video-streaming-service-cra- shes-during-usa-germany-world-cup- match-1201251221/ http://variety.com/2014/digital/news/ abcs-live-internet-oscar-stream-suffers- nationwide-outage-1201124215/
- 7. 7 vi. “Les nouvelles frontières de la diffusion OTT: Multicast & P2P” www. ovfsquad.fr vii. Cisco® 2014 Visual Networking Index viii. For more information on skewed video popularity distribution, see: “Watching Videos from Everywhere: a Study of the PPTV Mobile VoD System,” Social networks have a tendency to skew this distribution even more, to such an extent that a mere 2% of content can garner up to 90% of views. See “Video Requests from Online Social Networks: Characterization, Analysis and Generation,” http://www.cs.sfu. ca/~jcliu/Papers/VideoRequests.pdf PEER-ASSISTED DELIVERY: OPTIMIZE VIDEO PLAYOUT AND MINIMIZE INFRASTRUCTURE RISK P2P: A VIABLE SOLUTION To remain competitive, broadcasters will have to ensure high-qua- lity content at lightening-fast speeds – every time. They will have to be prepared to support an increasing number of visitors all while reducing their dependency on traditional delivery networks. Band- width is a video platform’s largest production cost, and those that manage to control this expense without compromising quality will have an edge up. CDNs admit that additional distribution models are necessary, and that time is of the essence. One of the largest actors cites peer- to-peer streaming as a solution that has already proven effective on a large scale.vi When combined with traditional unicast delive- ry, peer-to-peer adaptive streaming presents a viable alternative. Why? Let’s take a look at two realities: Peak-hour Internet traffic is growing faster than average traffic.vii Over 80% of video traffic is concentrated on 20% of content.viii The videos that are and will be in highest demand are also those being watched by a large number of viewers at the same time. Peer-to-peer streaming allows simultaneous users to exchange video segments among themselves rather than each connecting to a server to do so. The system thus effectively overcomes seve- ral of the scalability obstacles presented by the video market. First, it largely reduces broadcasters’ reliance on CDNs for content distribution, drastically reducing bandwidth costs, freeing up congestion and protecting them from infrastructure malfunctions. 2
- 8. 8 With a simple CDN-based model, users download directly from the server, using up bandwidth and creating congestion. With a hybrid peer-to-peer system, viewers can supply each other with the content, optimizing traffic flows and ensuring better QoS. Decentralizing this exchange of data also significantly improves the quality of service for the end user, as each viewer is able to collect the segment needed from the source that can provide it most quickly. Lastly, it turns inevitable user volume increases into an asset, as the solution works better the more viewers are watching the same content. THE STREAMROOT ANSWER StreamRoot has developed a hybrid peer-to-peer video streaming solution based on the latest Internet technologies: HTML5, JavaScript, Media Source Extensions and WebRTC. Unlike other peer-to-peer solutions, StreamRoot’s technology is transparent, requiring no plugin, extension or other installation on the part of end users. When a StreamRoot user accesses a webpage, the video content beginsloading directly fromtheserver.Atthesametime, theviewer connects to the StreamRoot tracker and retrieves an intelligently selected list of peers, establishes a direct connection with them and requests video segments. If the peers cannot provide the segments quickly enough, it automatically switches back to the origin/CDN server, guaranteeing at the very least the same quality of service as a CDN-only solution.
- 9. 9 Streamroot leverages two new cutting-edge technologies. The first, WebRTC, is a new standard that allows users to establish direct and secure real-time communications with other users without worrying about NATs and firewalls. It is included natively in browsers as a javascript API, and is available as an open-source library distributed by Google for easy integration into any type of device. The second is Media Source Extensions, another web standard designed for dynamic management of video streams directly in HTML5. Avoiding the need to rely on cumbersome Flash systems, MSE has already been adopted by most browsers, and is used by default by Netflix and YouTube. Like WebRTC, it is also becoming standard for mobile devices, set top boxes and smart TVs.ix Harnessing these technologies, our solution can be decomposed into three main modules: Media engine module This module enables adaptive bitrate streaming play- back in HTML5 and Flash. It supports the newest MPEG-DASH streaming standard, as well as older for- mats like Smooth Streaming and HLS. It uses dynamic adaptive streaming algorithms based on the end-user’s bandwidth to provide the best possible experience the viewer’s device and connection can offer at a given time. Peer-to-peer module This module enables direct peer-to-peer data trans- fer between viewers. It uses multiple proprietary StreamRoot algorithms to optimize exchanges in both Live and VOD playback modes. Format agnostic, it can be used with DRMs and integrated into a custom player via a media interface. Finally, it gathers several useful analytics on the viewer’s performances and behavior. Tracker Built on lightweight and scalable technologies like Node.JS and Redis, the tracker serves as a relay for es- tablishing peer-to-peer channels, dynamically selecting the best peers based on geographical and topological criteria. It also provides the security required by a pro- fessional video content distributor: geoIP and domain restriction, along with content integrity verification. ix. For more technical information on WebRTC and Media Source Extensions, consult the World Wide Web Consor- tium’s (W3C) Editor’s Drafts: http://dev. w3.org/2011/webrtc/editor/webrtc. html and https://dvcs.w3.org/hg/ html-media/raw-file/tip/media-source/ media-source. html.
- 10. 10 Today, StreamRoot provides the first and only workable peer- based alternative to unicast distribution that has been tested on a significant scale. The system is currently compatible with Internet browsers that have adopted WebRTC: Chrome, Firefox and Opera. Internet Explorerx and Safari are likely to follow suit, making the solution available on all major desktop systems. Mobile (Android and IOS) and Smart TV compatibility is currently under develop- ment and should be fully functional in 2015. P2P helps minimize infrastructure risks Despite their limitations, Content Delivery Networks are an important part of today’s media workflow. Peer as- sisted delivery can easily work in tandem with a CDN, optimizing the overall streaming experience, while at the same time reducing economic dependence on CDNs and ensuring that viewers can continue watching their content even if the CDN and local server expe- rience temporary outages. P2P cuts costs Initial tests show that bandwidth costs can be reduced by up to 70% with an effective peering solution in place. P2P optimizes video playout as demand rises Traffic peaks are no longer a source of stress but an advantage, as peer-to-peer systems work better the more peers there are available to share content, and can greatly improve the scalability of the streaming architecture. x. Microsoft co-authored the draft speci- fication for WebRTC 1.1.
- 11. 11 0 6 12 18 4:10 4:20 4:30 4:40 4:50 5:00 5:10 5:20 5:30 5:40 5:50 6:00 6:10 6:20 6:30 6:40 6:50 7:00 7:10 CDNP 2P CASE STUDY: OVER 50% PEER-TO-PEER STREAMING ON A LARGE VOD PLATFORM P2P: A VIABLE SOLUTION StreamRoot recently partnered with a Russian video-on-demand platform to test its peer-to-peer solution on the website’s most po- pular video during a Friday night peak. The following parameters were used: 120 minutes of video, in MPEG-DASH, h264/AAC, at 1,000 kbps a large geographical area spanning 6 Russian-speaking countries 2,089 users at peak, i.e. a total bandwidth of 2.1 Gbps or 1TB per hour This trial clearly demonstrated the scalability of the P2P model, as the more peers there were, the more data was transferred between them. In this trial, StreamRoot achieved up to 58% peering, with percen- tages rising at peak use times (42 to 58% streaming at traffic peaks). 3 Streaming traffic from 4 pm to 7:10 pm on the site's most popular video Time Volumestreamed(inGB)
- 12. 12 6:15 6:16 6:17 6:18 6:19 6:20 6:21 6:22 6:23 6:24 6:25 0 6 12 18In the middle of the test, the origin server went down entirely from 6:15 to 6:19 pm. P2P streaming ensured halfOftheserviceduringtheoutage–50%ofuserswere entirely unaware that the server had malfunctioned. This case study helped demonstrate: P2P requires a relatively low critical mass. For effective peering, the critical mass for a 120-minute video file (i.e. P2P > 30%) is less than 100 simultaneous viewers. For a 10-minute video, only 20 peers would be needed. P2P is an effective safeguard against server outages Peers ensured 50% of the service that would have been completely lost by the server outage. P2P optimizes video playout as demand rises The overall average download speeds are 14% higher with peer-assisted delivery, and efficiency improves greatly with geolocation and topology based prioritiza- tion algorithms. Time Volumestreamed(inGB) Momentary Server Outage
- 13. 13 FINAL REMARKS With exponential growth in viewer numbers, exploding file sizes and increasing use of mobile devices and tablets, the online video industry is poised for unprecedented demand. As recent outages have shown, even the most robust infrastructures are not ready to handle the influx of users that large broadcasters are beginning to experience today. Another model is needed in addition to traditional unicast distribution. Peer-to-peer adaptive streaming provides a viable, scalable supplement to traditional CDN distribution. With its ability to transform increasing viewer numbers into an asset, peer-assisted streaming based on the latest Internet technologies such as HTML5, Media Source Extensions and WebRTC can offer broadcasters a key competitive advantage, enabling them to reduce costs, improve streams and lighten the burden on saturated network infrastructures. 4
- 14. ABOUT STREAMROOT StreamRoot offers a solution for video streaming combining standard unicast delivery (e.g. CDN) and peer-to-peer protocols based on HTML5 and WebRTC. Founded in France in 2013 by three engineers from Ecole Centrale de Paris, StreamRoot participated the Le Camping and Techstars Boston accelerator programs and has been recognized with numerous awards including the Trophée Startups Numérique and Hello Tomorrow Challenge. At the leading edge of HTML5 adaptive streaming technology, StreamRoot created the first MPEG-DASH peer-assisted video player in HTML5 working for both Live and Video on Demand streaming, and has since expanded its expertise to other adaptive streaming formats and platforms. Using WebRTC, our Peer-to-Peer API creates an edge network made up of viewers, which relieves broadcasters’ server infrastructures and bandwidth without requiring any action from the end-user. Now headquartered in the United States, StreamRoot is currently expanding its customer base in America and confirming the efficiency of peer-assisted delivery with large-scale use cases. For more information or to try StreamRoot: contact@streamroot.io.