Performance evaluation of parallel download in streaming services

In the Internet, video streaming services in which users can enjoy videos at home are becoming popular. In particular, video streaming with HDTV or UHDV class quality is expected to attract many users in the future. However, the transmission bit rate of HDTV or UHDV is quite large, so traffic flows generated by this service will severely affect other flows. To alleviate the impact caused by this bandwidth-demanded flow, mitigating the traffic concentration and balancing the load on links in the network is important. One simple approach to achieving this goal is to provide multiple video servers over the network and download video data from multiple servers to one user. In this paper, we numerically evaluate the effectiveness of parallel download using 23 topologies of commercial ISP networks. We demonstrate that the effectiveness of parallel downloading strongly depends on the network topology. We also show required conditions to effectively decrease the variation in link load without increasing the average link load.