Over the past few years, the Dynamic Adaptive Streaming over HTTP (DASH) standard has been widely adopted by video streaming services; this has led to a considerable amount of research on efficient adaptation algorithms to maximize users’ Quality of Experience (QoE), but the comparison between algorithms is often flawed and unrealistic. Trace-based simulations or experiments in wired testbeds often fail to capture the full complexity and variability of a live network environment, and the comparison between them is often biased by the simplistic assumptions made about the simulation/testing setup. In this work, we implement four of the most representative adaptation algorithms and compare their performance in a real campus network. This allows us to have a fair comparison of the various algorithms’ strengths and weaknesses in a representative scenario, improving the understanding of the dynamics of video streaming adaptation and highlighting the open problems in the field. The test results make a clear trade-off emerge, as algorithms designed for high fairness in a static scenario cannot deal with a dynamic channel and cross-traffic effectively and result in very low QoE levels, while more aggressive algorithms face rebuffering events in fast-varying scenarios.
Comparing DASH adaptation algorithms in a real network environment
Chiariotti F.;Gadaleta M.;Zanella A.;Rossi M.
2019
Abstract
Over the past few years, the Dynamic Adaptive Streaming over HTTP (DASH) standard has been widely adopted by video streaming services; this has led to a considerable amount of research on efficient adaptation algorithms to maximize users’ Quality of Experience (QoE), but the comparison between algorithms is often flawed and unrealistic. Trace-based simulations or experiments in wired testbeds often fail to capture the full complexity and variability of a live network environment, and the comparison between them is often biased by the simplistic assumptions made about the simulation/testing setup. In this work, we implement four of the most representative adaptation algorithms and compare their performance in a real campus network. This allows us to have a fair comparison of the various algorithms’ strengths and weaknesses in a representative scenario, improving the understanding of the dynamics of video streaming adaptation and highlighting the open problems in the field. The test results make a clear trade-off emerge, as algorithms designed for high fairness in a static scenario cannot deal with a dynamic channel and cross-traffic effectively and result in very low QoE levels, while more aggressive algorithms face rebuffering events in fast-varying scenarios.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.