Real-time Streaming Reliability and Performance Optimization Using Content Steering

Media streaming over the Internet is dominated by HTTP adaptive streaming solutions such as DASH and HLS. These solutions typically host the required manifest files and media segments on multiple content delivery networks (CDNs). To avoid outages and performance problems and for scalability reasons, selecting the optimal CDN is an essential step in the media streaming workflow. Content steering provides a deterministic capability for a content distributor to switch a player's CDN source, both at startup and in the middle of a session, through a remote out-of-band steering service. This paper focuses on the theoretical and practical aspects of content steering for DASH. We highlight the important features of the DASH-IF Content Steering specification and demonstrate our implementation in an open-source media player (dash.js). Moreover, we illustrate how streaming metrics based on the CTA-5004 (Common Media Client Data, CMCD) specification and information about the Internet service provider can drive the selection mechanism on the steering server. Our results show that content steering enables a seamless transition between different CDNs during a running streaming session and has the potential to significantly improve the quality of experience for the end-user.