On the Interaction of Adaptive Video Streaming with Content-Centric Networking

TL;DR

This paper investigates how adaptive HTTP video streaming interacts with in-network caching, revealing challenges like degraded cache hit rates and oscillations, and proposes DASH-INC for improved streaming performance.

Contribution

It provides an experimental and analytical study of adaptive streaming over caching networks, highlighting issues and introducing the DASH-INC framework for enhancement.

Findings

Cache hit rate significantly decreases with adaptive streaming.

Frequent cache-server oscillations occur due to throughput misjudgments.

In-network caching impacts rate adaptation effectiveness.

Abstract

Two main trends in today's internet are of major interest for video streaming services: most content delivery platforms coincide towards using adaptive video streaming over HTTP and new network architectures allowing caching at intermediate points within the network. We investigate one of the most popular streaming service in terms of rate adaptation and opportunistic caching. Our experimental study shows that the streaming client's rate selection trajectory, i.e., the set of selected segments of varied bit rates which constitute a complete video, is not repetitive across separate downloads. Also, the involvement of caching could lead to frequent alternation between cache and server when serving back client's requests for video segments. These observations warrant cautions for rate adaption algorithm design and trigger our analysis to characterize the performance of in-network caching for HTTP streaming. Our analytic results show: (i) a significant degradation of cache hit rate for adaptive streaming, with a typical file popularity distribution in nowadays internet; (ii) as a result of the (usually) higher throughput at the client-cache connection compared to client-server one, cache-server oscillations due to misjudgments of the rate adaptation algorithm occur. Finally, we introduce DASH-INC, a framework for improved video streaming in caching networks including transcoding and multiple throughput estimation.