Broadcasting video with the knowledge of user delay preference

In designing a video broadcasting system, the delay preference of a user is traditionally regarded as unknown. In fact, such preference can be known upon user\´s arrival by employing some techniques such as: (i) delay-dependent charging, where users are offered different levels of pay-per-view (PPV) depending on the maximum delay they are willing to tolerate; or (ii) reservation, where a user specifies the exact play-time of a movie in advance, and he/she is charged according to the length of the reservation period. We explore, for the first time, the impact of such delay knowledge on request scheduling and system cost in terms of user loss and stream requirement. For delay-dependent charging, we propose "delay-aware broadcasting" (DAB) and its variant based on reservation (DAB-r), where allocation of server streams is driven by the delay tolerance of a user. DAB-r offers differentiated grade of services according to user PPV (and thereof classes). As compared with a system where user delay preference is not known, our schemes achieve substantially lower user loss rate, higher revenue, and better fairness. Regarding reservation system, we consider a scheme where clients can pre-buffer video data. Unicast streams are used to merge requests back to the on-going broadcast streams. We show that a reservation system achieves substantially lower stream requirement as compared to an on-demand system based on "patching".