Cross-layer metrics sharing for QUICker video streaming

Abstract

QUIC is marketed to hold many advantages over TCP. However, preliminary experimentation has shown that simply running contemporary HTTP Adaptive Streaming (HAS) implementations over QUIC does not improve but actually hurts streaming performance compared to a traditional TCP deployment. We argue that this behavior can be attributed to the amount of TCP specialization that HAS Adaptive BitRate (ABR) algorithms have received over the years. In contrast to TCP (which can be regarded as a "black box"), QUIC actually encompasses all the necessary tools to empower streaming performance optimization (e.g., definition of custom congestion control algorithms, access to transport-layer metrics in the application layer). This however comes at the expense of added complexity which in turn could lead to misinterpretations of the root causes of suboptimal streaming performance. To facilitate research on HTTP adaptive bitrate streaming over QUIC, in this paper we propose a solution towards jointly visualizing transport-and application-layer metrics to allow for a better understanding of HAS streaming performance over various types of transport protocols (i.e., TCP versus QUIC). We see the work presented in this paper as an important stepping stone towards cross-layer optimization of HAS ABR performance to achieve a better overall Quality of Experience (QoE) for streaming users.