Evaluating DASH Player QoE with MPTCP in Presence of a Shared Bottleneck Link

Improving Quality of Experience for video streaming remains a challenge as the Internet consists of numerous shared links which turn into shared bottlenecks due to lack of Quality of Service measures or inconsistent QoS policies between ISPs. Solutions to this have come up involving Dynamic Adaptive Streaming over HTTP (DASH) (that adapts its playback bitrate according to available throughput) with Multipath TCP (that uses multiple network interface paths to increase throughput and reliability) as transport protocol. Even then, DASH QoE can degrade as MPTCP subflows between same end-system pairs can overlap on shared bottleneck links. To reduce this degradation, we need to understand the behaviour of DASH on MPTCP in that scenario. While the current MPTCP literature has extensive coverage on shared bottleneck links in congestion control design and bandwidth usage fairness, the same is not true for DASH QoE. We rectify this shortcoming by presenting a detailed comparative performance evaluation between DASH on MPTCP and DASH on TCP in presence of a shared bottleneck link in terms of objective QoE parameters under various video segment sizes with the results from a topology having independent links as a reference. There are two major outcomes of this experimental evaluation. First, the optimal video segment size for best QoE is a value large enough to incur least network overhead while being small enough to not congest the shared bottleneck. Second, it is verified that the objective QoE parameters have strong correlation to network resource allocation and ISPs can use them to improve real-time video streaming on their network.