Measuring temporal redundancy in sequences of video requests in a News-on-Demand service

Abstract Streaming media is becoming one of the major components of Internet traffic. Therefore, a better understanding of users’ video request patterns is essential, in order to design an effective and efficient video distribution system (caching, storage capacity, bandwidth, etc.). In this paper, the core issue will be the analysis and modeling of video requests temporal redundancy. The study will be centered on a News-on-Demand (NoD) service, which provides support to a wide variety of digital newspaper editions from different regions of Spain. Specifically, six digital newspapers with a high number of requests were analyzed during a period of one year. The level of redundancy has been measured by a global (gR) and a partial redundancy (pR) method, which is new in this type of services. As a result, the main contribution of our paper is a global and partial redundancy model for each digital newspaper, which would allow us to forecast the level of video requests likely to be repeated in the near future. The model turned out to be user independent and with a timeless effect. The validation process shows that all the models successfully pass the hypothesis test, which means that there were no significant differences between the model and the real data. The pR models could predict between 1% and 6% of video requests temporal redundancy with a level of accuracy which varies between 88% and 100%.