Audience Aware Streaming

Live linear streams are typically produced using a fixed bitrate ladder and codec mix. This approach locks in storage, compute, delivery costs and power consumption. Unpopular channels have the same operating costs as popular ones. Content distributors would much rather focus their resources on their popular content and maximize aggregate quality of experience (QoE) while minimizing storage, compute, delivery and power. This paper describes a collaboration between Ateme (a provider of video compression and delivery solutions) and Akamai (a global compute and delivery provider) in which such an optimization is achieved. Player viewing sessions are collected by Akamai using the common media client data (CMCD) standard and fed in a real-time data feed to an Ateme Stream Optimizer. The Optimizer gathers data on the viewership count, geo diversity, device diversity and device capability and then dynamically adjusts the cloud compute resources allocated to the stream, varying the bitrate ladder and codec mix offered so as to provide the best quality to the popular streams while minimizing delivery costs and power consumption The paper describes the proof-of-concept system that was built, and presents the real-world results that were obtained to validate the hypothesis that live linear OTT operations can be optimized with the addition of real-time audience data.

Mickaël Raulet | ATEME | France
Josselin Cozanet | ATEME | United States
Khaled Jerbi | ATEME | Canada
Will Law | Akamai | Switzerland