Achieving High Utilization by Elastic Chunk Scheduling in DASH Systems

Dynamic adaptive streaming over HTTP (DASH) has been widely deployed to provide various video services in the Internet. However, the HTTP/1.1 or HTTP/2 utilized by the DASH system cannot ensure high quality of user experience in highly dynamic network scenarios. Specifically, when the clients fetch the video chunks from servers, it is well known that HTTP/1.1 protocol suffers from low utilization due to its stop-wait fashion. Though HTTP/2 enables the servers proactively push multiple chunks to clients on a single request, it exhibits poor adaptability to the network dynamic, since the bitrate of multiple chunks is fixed in one push cycle. To address these inefficiencies, we propose elastic chunk scheduling (ECS) to adaptively adjust the batch size of chunks in each push cycle according to network dynamic. To achieve high utilization, ECS increases the batch size of chunks under low network dynamic. Otherwise, ECS reduces the batch size to flexibly change the bitrate of chunks to avoid rebuffer and overflow at the player buffer. Through achieving good tradeoff between high link utilization and flexible rate switching, ECS improves overall QoE performance. The experimental results of testbed implementations show that ECS greatly decreases the rebuffer ratio by up to 36% and increases the average bitrate by up to 10% compared with the state-of-the-art solutions.