A cache-level quality of experience metric to characterize ICNs for adaptive streaming

Abstract

Adaptive streaming has motivated Information- Centric Network (ICN) designs to improve end-user quality of experience (QoE). However, their management and evaluation rely either on conventional cache-level metrics that are poor representations of QoE, or consumer-side indicators that are opaque to network services. This paper proposes a measure to bridge the gap between cache performance and consumer QoE. We introduce Maximal Sustainable Bitrate (MSB), defined as the highest bitrate deliverable in time to be in time to meet a given request without buffering. Based on our observations, we posit that QoE is maximal when requested bitrates match a cache’s MSB for that content. We design a cache-level reward function as a benchmark metric that measures the difference between requested bitrates and MSB. We hypothesize that aggregated rewards are an indicator of overall system performance. Performance evaluations show high correlation between the sum of cache rewards and consumer QoE.