Dynamic Segment Size Selection in HTTP Based Adaptive Video Streaming

Video streaming takes the lion's share of network bandwidth, with a trend that will likely increase in the future. In the recent years, dynamic adaptive streaming has been developed to offer a smooth video stream with variable quality, depending on the performance of the network connection. At present, several content providers like Netflix, YouTube and Hulu, to name a few, already offer videos that can be streamed with a dynamic video quality. In this work we study the tradeoff of the video length to be downloaded, called segment size, and show the differences in terms of playback quality and reduced buffer outages. We then propose an algorithm able to exploit the network conditions, and adaptively select the video segment size to be downloaded. We analyze the benefits of our proposal in terms of increased playback quality and reduced buffer outages against classic fixed segment size solutions. Our result show that dynamically adapting the segment size can reduce buffer outages, while also increasing the quality streamed.