A Comprehensive Queuing Model For Adaptive Video Streaming
Join Zoom Meeting https://nus-sg.zoom.us/j/91562142721?pwd=dUNyMHJmOUFSZXhMeUZLRDVFVVlPdz09
Abstract:
Dynamic Adaptive Streaming over HTTP (DASH) relies on a rate adaptation component to decide on which representation to download for each video segment. A plethora of rate adaptation algorithms has been proposed in recent years. The decisions of which bitrate to download made by these algorithms largely depend on several factors: estimated network throughput, buffer occupancy, and buffer capacity. However, these algorithms are not informed by a fundamental relationship between these factors and the chosen bitrate. As a result, we found that they do not perform consistently in all scenarios, and require parameter tuning to work well under different buffer capacity. In this work, we model a DASH client as an M/D/1/K queue, which allows us to calculate the expected buffer occupancy given a bitrate choice, network throughput, and buffer capacity. Using this model, we propose a simple rate adaptation algorithm QUETRA (QUEuing Theory-based Rate Adaptation). We then extend the QUETRA into three parts. First, we develop a Distributed Queuing theory-based rate adaptation algorithm for DASH (DQ-DASH) for multi-server streaming that models the DASH client downloading from multiple servers as an Mx/D/1/K queuing system. DQ-DASH facilitates the aggregation of bandwidth from different servers and increases fault-tolerance and robustness through path diversity. Second, we use QUETRA for 360-degree tiled video streaming, where the DASH client mix the tiles from different representations during playback based on the Field of View (FoV). We develop a Multiclass-Knapsack-based method to allocate bitrate to the tiles based on the FoV using a dynamic value calculation for the item in the class of the knapsack. Finally, we extended QUETRA for formulating rate and playback speed adaptation to achieve low latency for live video streaming. We also propose a simple and intuitive method for bandwidth estimation in low-latency live video streaming. The method analyzes the segment's chunk and the Maximum Transfer Unit (MTU) of the communication medium to achieve consistent accuracy regardless of segment duration.
We evaluated QUETRA and its extensions under a diverse set of scenarios and found that, despite its simplicity, its performance is much higher. QUETRA leads to a better Quality-of-Experience (QoE) (7%-140%) than existing rate adaptation algorithms, and the DQ-DASH improves the QoE up to 27% compared to the best performing least-connection CDN load balancing rule. QUETRA for 360-degree tiles video streaming with proposed Multiclass-Knapsack-based bitrate allocation reduces both the stall duration by up to 43% and the bandwidth wastage for non- visible tiles by 41%. The rate and playback speed adaptation extension of QUETRA controls the latency within limits as small as 1 second without causing excessive speed up. It also plays the video at a normal speed for more than 87% of the total playback time without skipping any chunks.