Understanding and decreasing the network footprint of catch-up tv

摘要: "Catch-up", or on-demand access of previously broadcast TV content over the public Internet, constitutes a significant fraction peak time network traffic. This paper analyses consumption patterns nearly 6 million users nationwide deployment catch-up service, to understand support required. We find that has certain natural scaling properties compared traditional TV: The nature spreads load time, and have much higher completion rates for streams than reported. Users exhibit strong preferences serialised content, specific genres.Exploiting this, we design Speculative Content Offloading Recording Engine (SCORE) predictively records personalised set shows on user-local storage, thereby offloads traffic might result from subsequent access. Evaluations show even with modest storage ~32GB, an oracle complete knowledge user can save up 74% energy, 97% bandwidth current IP streaming-based architecture. In best case, optimising energy consumption, SCORE recover more 60% savings achieved by oracle. Optimising purely rather reduce bandwith additional 5%.