Control challenges in multi-level adaptive video streaming

Streaming video is one of the fastest-growing applications of the Internet. The Internet's diversity and dynamism demands that video streams adapt to ensure maximum quality at all times. This paper describes the control challenges we have encountered in the Quasar project's "multi-level" adaptive streaming video player. We first describe the framework and environment of the player. This framework uses software feedback to control resource allocation as well as the quality of media delivery. We present the control challenges raised by our framework which include horizontal and vertical feedback composition, difficult to model systems, and unpredictable, nonlinear actuators. We describe some of the approaches we are taking to address these challenges, related work, and future application areas and the challenges they will raise.