Advancing Multimedia Distribution vCache : Supporting Cost-Efficient Adaptive Bitrate Streaming

Guanyu Gao,Yonggang Wen,Jianfei Cai

2017-01-01

Abstract:V ideo streaming dominates Internet traffic, accounting for more than 70 percent of North American downstream traffic at peak time.1 However, limited bandwidth capacity, unstable network conditions, and diverse viewing devices inherently deteriorate user experiences, triggering a tussle between the growing demand for video traffic and the quality of viewing experiences.2 Adaptive bitrate (ABR) streaming is a widely adopted solution for improving viewing experiences under such conditions. Figure 1a illustrates the video processing flow for ABR streaming. Each video must be transcoded into multiple representations and then cached in streaming servers. A media presentation description (MPD) file is required to manifest the available representations for a video.3 When starting a video session, the video player first obtains the MPD file and then selects the best possible quality representations according to the current network conditions and device capacity. However, transcoding is compute-intensive and consumes tremendous resources. Caching multiple representations of a video requires several times more storage space than just caching one representation. Thus, ABR streaming of videos can greatly increase operational costs. In contrast to the tremendous resource consumption of ABR streaming, only a small percentage of video chunks are frequently requested by users. Specifically, the top 10 percent of the most popular videos account for almost 80 percent of total views.4,5 Furthermore, for 60 percent of video sessions, less than 20 percent of the durations are viewed; most users abort viewing within 40 seconds.6,7 These user viewing patterns reveal that users consume only a small fraction of video chunks. Considering the tremendous computing and storage resource consumption required to transcode and cache videos, is it necessary to pre-transcode each video and cache all video chunks? To answer this question, we designed vCache, a network function virtualization (NFV)-based virtual caching scheme for managing videos for ABR streaming to minimize overall operational costs.