Cross Video HTTP Adaptive Streaming for Short Video Improvement

HTTP adaptive streaming (HAS) has been successfully used by many companies (e.g. YouTube, Netflix, etc.) to optimize their streaming experience. However, it also brings some unique challenges. For example, HAS doesn't work well for YouTube's videos that only last a minute or two. According to the experiment results, the HAS client will take tens of seconds to switch to the high quality stream, which degrades the experience for viewers when they are watching the short term videos. To resolve this issue, we propose a method for network assistant cross video HTTP adaptive streaming (NC-HAS), which has several features: (1) The data rate adaptation is carried out across two successive videos rather than in one video by utilizing the fact that clients often watch more than one short video continuously; (2) The network nodes monitor the network traffic to avoid the performance degradation caused by the traffic burst during the interval between two successive videos; (3) The client transparent scheme avoids the modification in massive HAS clients. The experiments results show that the time to achieve the high quality stream at the client is decreased greatly by adopting NC-HAS. At the same time, the user's throughput and the cell utilization is improved, which can be regarded as the byproduct of NC-HAS.