Predictive network traffic engineering for streaming video service

Next Generation Network services with Fiber-to-the-home have been spreading in Japan; therefore, the number of viewers using Video on Demand (VOD) services has also increased. Network operators are required to maintain service quality during peak hours, so they need to design bandwidth for VOD at those times. It is obvious that congestion occurs when large number of viewers watch VOD programs simultaneously. However, viewer behavior does not depend on only internal factors of VOD service but also external ones such as natural phenomena and social events. This makes it difficult for network operators to predict how traffic will increase and to design bandwidth adequately. We analyzed actual clarified that it is effective to use a log-normal VOD traffic and distribution model. We also predicted traffic distribution based o n Bayes' method.