Adaptation method for video streaming over HTTP/2

Abstract

HTTP streaming, which is currently based on the pull-based HTTP/1.1 protocol, has a tradeoff problem between overhead and adaptivity. We propose an adaptation method for adaptive streaming over the new HTTP/2 protocol, leveraging its server push feature. The method is based on a cost function that takes into account the number of pushed segments and the client buffer level. Experiment results show that the proposed method can improve the balance between the number of requests and buffer stability compared to existing methods.

Seamless Mobile Video Streaming over HTTP/2 with Gradual Quality Transitions

“…It should be noted that the existing methods strictly decide the number of pushed segment in advance (e.g., [7]) or based on buffer level only (e.g., [10]). Meanwhile, our method flexibly makes decisions based on both buffer level and connection throughput.…”

“…In this section, we evaluate our method using a simple bandwidth trace and a complex bandwidth trace obtained from a mobile network. For reference methods, we use the Push-N method used in [4], [7] and the request-buffer-based (RBB) method proposed in [10]. In the former method, N segments are sent for each request.…”

“…In the initial buffering stage, the simple Push-N method with N = 1 is employed. The testbed of our experiments is similar to that of [10], which includes an HTTP/2 web server, an HTTP/2 client and an IP network. The IP network includes a router and wired connections connecting the server and the client.…”

“…It should be noted that in mobile networks, as the throughput is fluctuating, the above methods may result in large bitrate reductions and such rate-switching events definitely reduce the quality of service [12]. In this paper, we extend our preliminary work in [10] and additionally consider bitrate variations. Instead of selecting an adaptation pair (R, N ) of bitrate R and number of pushed segments N for only one next request, we decide adaptation pairs for some future requests.…”

“…In the literature, the server push feature with a fixed number of pushed segments per request has been investigated for low-delay streaming [7], low request-related overhead [4] and power efficient mobile streaming [8]. Among existing studies, our previous studies in [9], [10] are the first ones that raise the need to adaptively decide the number of pushed/requested segments in each request to balance the requirements of low request-related overhead and buffer stability. In [11], the number of pushed segments is adapted to avoid the "over-push" problem, in which network resources are wasted when a user decides to stop watching a video after checking the first few seconds.…”