Evaluating DASH QoE with MPTCP Under Different MPTCP Buffer Sizes and Path Latencies

Abstract

Dynamic Adaptive Streaming over HTTP (DASH) is accessible without any restrictions on most networks and can adapt playback bitrate to match available throughput, which ensures acceptable Quality of Experience (QoE) to content consumers. When DASH is used with MPTCP as transport protocol, the aggregated throughput and increased reliability due to the use of multiple interfaces for a streaming session improves QoE. There are ongoing efforts to improve QoE even further. The current literature concerning DASH on MPTCP focuses on studies related to link errors, shared bottleneck links and improvement of QoE using player level approaches. Moreover, the experimental results in current literature are obtained from software-based simulators or emulators utilizing very simple topologies which may not be consistent with real-world usage. Our work is the first evaluation of DASH QoE performance on MPTCP under different MPTCP buffer sizes and path latencies. Our work is also the first to use a real hardware testbed with a generalized Internet topology for DASH on MPTCP performance evaluation. There are four major outcomes of our experimental evaluation. First, the optimal MPTCP buffer size for the best end-user QoE is dependent on the bandwidth-delay product within a reasonable path latency limit. Second, the path latency and the video segment size that optimises QoE increase proportionally. Third, beyond the path latency limit, DASH is non-functional due to problems with HTTP/1.1 head-of-line blocking and MPTCP congestion control. Last but not the least, the MPTCP buffer size does not influence raw network throughput directly.