Scheduling Techniques for Media-on-Demand

Broadcasting popular media to clients is the ultimate scalable solution for media–on–demand. Recently, it was shown that if clients can receive data at a rate faster than what they need for playback and if they can store later parts of the media in their buffers, then much higher scalability may be obtained. This paper addresses scheduling problems arising from these new systems for media–on–demand. For given amount of bandwidth, we reduce the maximal start–up delay time for an uninterrupted playback. We achieve our results by introducing two techniques. In the first, the media is arranged on the channels such that clients gain from buffering later parts of the transmission before the actual start of the playback. In the second, segments of different media may be mixed together on the same channel. We introduce a simple class of recursive round–robin scheduling algorithms that implement both techniques. Our results improve the best known asymptotic results. Moreover, our scheduling algorithms outperform known results for practical values for number of media and number of broadcasting channels. For some specific small values, we present solutions that are better than those achieved by our algorithms. Finally, we show that our techniques are useful for models in which clients may not receive data from all the channels, and are applicable to media with different lengths and popularities.