Traffic prediction method based on semi-Markov process
基于半马尔柯夫过程的流量预测方法

a traffic prediction method based on semi-Markov process was presented. The semi-Markov process was used to describe network traffic characteristics and divide network traffic into four states: busy, idle, rising and falling. The prediction algorithm about traffic upper bound in busy state was concluded according to traffic stochastic distribution in each state and mutual transition relationships between states. Some practical trace data from WAN and LAN were employed to verify the traffic prediction method. About 95% trace data really follow the corresponding stochastic distribution in each state and 90% of predictions are correct with the probability of 0.8 or 0.9, and the relative error is lower than 15% for backbone networks.