Packet scheduling based on Geo^Y/G/∞ input process modeling for streaming video

We introduce a new stochastic process called Geo^Y/G/∞ input process with beta-distributed batch size and Weibull-like autocorrelation function in order to model video traffic. Investigating the overflow probability of a queueing system by large deviation theory, we develop a streaming scheduling algorithm by applying the overflow analysis result to estimating the packet deadline-missing probability. Through experiments with 30-minute long movie traces, we show that our proposed scheduling scheme outperforms existing schemes based on fractional Brownian motion and Markovian models.