Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Speed5G Workshop London presentation of 5G XCast

68 views

Published on

Presentation of the 5G XCast project at the Speed5G Workshop on March 7th 2018 in London

Published in: Engineering
  • Be the first to comment

  • Be the first to like this

Speed5G Workshop London presentation of 5G XCast

  1. 1. Rory Turnbull British Telecommunications 5G-Xcast Unified Content Delivery on Fixed and Mobile Networks A view of synchronous TV delivery 118
  2. 2. 119 Observations on content consumption
  3. 3. 120 Linear TV in decline Active viewing hours of on- demand vs live and scheduled linear TV
  4. 4. 121 Live audiences are very dynamic Traffic volumes over the EE network during the England vs. Wales football match during Euro16 compared with the previous day.
  5. 5. 122 “Appointment to view” broadcast audiences are very dynamic
  6. 6. Appointment to view VoD has a similar profile to live and linear with a huge spike in demand when initially released Consumers want to watch together and engage in social media commentary and Traffic profile would suit carousel broadcast and/or push prepositioning And VOD too…
  7. 7. 124 • Concurrent viewing of popular events drives traffic volumes – It doesn’t just change what people are watching, it drives overall demand volumes • This creates a capacity planning challenge • Broadcast and multicast can help manage peaks and simplify capacity management Observations
  8. 8. 125 How to deliver event-based content?
  9. 9. Traditional Broadcast Internet Two worlds • Only supports TV • Optimised network for national coverage of popular content • Highly efficient use of spectrum for simultaneous delivery at edge of network • High barrier to provisioning a new service • Only support linear delivery • One-size fits all • Supports many services • Non-optimised network for global coverage • Unicast at edge of network inefficient for simultaneous delivery • Very low barrier to provisioning a new service • Supports linear and on-demand • Highly customisable
  10. 10. Live On demand Delivery 127 The rise of the global platforms
  11. 11. 128 Content Delivery Networks Content Service Provider Content Delivery Network Network Service Provider End User
  12. 12. 129 • Generic technology is good • Non-specialist commoditised servers • Readily passes through firewalls • Don’t require specialist software or licences • It’s the cheapest option! Global Platforms Use HTTP
  13. 13. 130 But… HTTP Sub-optimal Video Streaming HTTP Timely Delivery Bitrate to match media Delivery Times Highly Unpredictable Throughput ignores media requirements What went wrong? Why not UDP + QoS?
  14. 14. Organisation B 131 Cross-organisational resource reservation is challenging Organisation A Latency Priority Capacity SLA Billing Resource Accounting Commercial Relationship Data / Content
  15. 15. Organisation B 132 Prefer Autonomous Resource Allocation Organisation A Resource Data / Content Internal Optimisation
  16. 16. Buffer 133 Clever apps not clever networks Clever Network Clever App Throughput Variation Guaranteed Bandwidth Adaptive Bitrate Streaming Network Handover Complex Network Handover Streaming Buffer Management ABR
  17. 17. 134 NSP needs capacity for many unicast streams Content Service Provider Content Delivery Network Network Service Provider End User Each user served their own unicast stream causing duplications and unnecessary load
  18. 18. 135 Best of both: CDN for global - dynamic selection of multicast at edge Content Service Provider Content Delivery Network Network Service Provider End User Multicast or broadcast can be used to save capacity when streams are the same
  19. 19. 136 • Combine CDN for global reach with multicast/broadcast for edge optimisation • Multicast/broadcast as internal network optimisation, rather than service to be sold • Servers and client applications work with unicast with standard Internet protocols (HTTP) • Application layer intelligence preferred over network signalling Content Delivery Framework Design Principles
  20. 20. 137 Unified Content Delivery - Challenges
  21. 21. 138 Synchronous Delivery of Asynchronous requests 1 2 3 4 Edge Node Time Independent, asynchronous HTTP requests usually responded to individually Aggregate requests and serve a single response
  22. 22. 139 • How do we keep end to end delay low enough for live? • How do we make it work with ABR? Quality control CDN delays UC/MC conversion MC/UC conversion Streaming buffer
  23. 23. 140 • How do will it work when the content and/or transport is encrypted? • Need to avoid having visibility of content internals Security and Trust CDN delays UC/MC conversion MC/UC conversion Streaming buffer
  24. 24. 141 Standards activity Significant update to Multicast operation on Demand (MooD) in Release 14 IP Multicast Adaptive Bitrate Developing Multicast ABR standard Relevant IP standards (e.g. media encapsulation, HTTP(S), multicast etc.)
  25. 25. 142 • We can combine global CDNs with multicast and broadcast at the edge of the network to get the best of both worlds • To make multicast and broadcast easy products to use, it should be possible to treat them as an internal optimisation capability, rather than a service to be sold • We should beware trying to “add value” to the network by over- complicating its APIs. The key takeaways
  26. 26. Any Questions ? 143

×