Where do you watch? A spatial analysis of online video service in mobile network

We present a specialized characterization work in the spatial dimension of online video service in mobile network. A better understanding of the spatial characteristics of such service is a key capability for network operators and service providers, but remains an unsolved problem. It is extremely challenging because the complexity to get the location information of the mobile users. In this paper, we address this problem head on by collecting a large-scale dataset containing over 9 billion network traffic traces from a cellular network in a north-eastern city in China. We further extract the video requests from those data. The user location information is cast into the information of cells visited by the users in the cellular network when they watch mobile videos. Based on this dataset, we provide a comprehensive characterization of the online video service in mobile network in the spatial dimension: 1) The overall geographic distribution of video requests is uncovered and hot spot regions are discovered. And the video request location entropy is defined to analyze how request time and video category impact the overall geographic distribution. 2) Then, the user mobility pattern is defined in terms of cell sequence and residence time. The movement of each mobile user while watching videos in mobile network and the overlapping of request locations between different sessions and different days of each user are further studied. 3) At last, the video request spatial distribution of each video is analyzed. The linear relationship between the video request spatial distribution and the video popularity is uncovered. In addition, the categories and user mobility patterns for the widely watched videos are analyzed. To the best of our knowledge, this is the first specialized study to shed light on the online video service in mobile network from the spatial angle. Our analysis provides a unique insight into how Youku is utilized by mobile users. We believe our observation findings can provide direct help to both network operators and service providers in practical.