Streaming - improving delivery efficiency

Streaming - improving delivery efficiency

Moderator: Darren Fawcett, VP Hardware and Systems Engineering - CommScope


Steve Appleby, Senior Manager, Research, Video Delivery - BT

Yuriy Reznik, VP, Research - Brightcove, Inc.

Multicast has promised to solve the scale issue of delivering very popular television over the internet for a long time. In our first paper from BT Research, both a proof-of-concept architecture and field trial results are presented for a means to assist unicast delivery using multicast. The paper addresses the challenges and needs arising from the requirement to support multiple different content service providers and their end client devices. Our second paper from Brightcove, builds on the recent addition of content steering into HLS and MPEG DASH (which enables dynamic routing of streaming content between different CDNs). This requires content steering servers and the challenge is how to implement them robustly and at scale. This paper discusses those challenges, presents a solution and an open-source project for validation and trial implementation. This session also has two supporting papers; the University of Surrey has very effectively used machine learning to expedite “elephant flow” detection within the network, enabling re-routing of this traffic to minimise latency induced congestion. The other is an investigation by Comcast into an emerging transport protocol, Media over QUIC, offering traffic prioritization and low latencies, without the typical short-comings of TCP-IP. Interesting experimental results are provided.