作者: Steven Weber
DOI: 10.1007/978-3-540-72606-7_56
关键词: Scale (chemistry) 、 Distributed computing 、 Process (computing) 、 Markov chain 、 Computer science 、 Expected value 、 Transfer (computing) 、 The Internet 、 Hitting time 、 Duration (project management) 、 Mathematical optimization
摘要: This paper addresses a practical problem in our everyday use of streaming media on the Internet: as user observes buffering stream with an uncertain transfer rate, when should that initiate playback stream? The tension is initiating prematurely will increase likelihood buffer starvation, while delay undesirable because it necessitates waiting. Three policies are studied: optimal policy (exploiting full knowledge process), static (the expected value policy), and online assuming only rate observed thus far. Lower upper bounds derived well associated minimum cost; these expressed terms (random) hitting time process. Simulation results for Markov modulated process identify near-optimal depending scale duration stream.