Understanding the Achieved Rate Multiplication Effect in FlowQueue-based AQM Bottleneck
Jonathan Kua · 2021
The progressive adoption of Active Queue Management (AQM) and the popularity of Dynamic Adaptive Streaming over HTTP (DASH)-based streaming services have motivated the development of adaptive chunklets. Chunklets significantly improves the Quality of Experience (QoE) of video streaming applications in the presence of cross-traffic, which is known as the Achieved Rate (AR) multiplication effect. However, the detailed behaviour of chunklets and their implications on queuing dynamics have not been well-explored. In this paper, we instrumented a fine-grained, packet-driven FreeBSD queue measurement kernel module, and present an experimentally validated system model to help us better understand the AR multiplication effect in FlowQueue-based AQM bottlenecks. We experimentally demonstrated the accuracy of our system model across a wide range of network settings, along with providing fine-grained insights into queuing behaviours, including the detection of hash collisions. Our approach can provide network operators and researchers with greater network visibility where AQM schemes are deployed.