VLSC: Video Live Streaming Using Cloud Services

Abstract

Live streaming of video contents over the Internet generally requires conversion (i.e.,transcoding) of the video contents based on the characteristics of viewers' devices (e.g.,spatial resolution, network bandwidth, and supported codec). Due to the complexity of video transcoding process, live streaming service providers are becoming reliant on cloud services (e.g.,Amazon AWS1). With the scalable and reliable processing capability that clouds offer, live streaming service providers are able to transcode the live streams in a timely manner and fulfill viewers' Quality of Service (QoS) demands. For that purpose, cloud services must be utilized efficiently. In this paper, we present a cloud-based architecture that facilitates transcoding for live video streaming. Then, we propose a scheduling method for the architecture that is cost-efficient and satisfies viewers' QoS demands. We also propose a method -- utilized by the scheduler -- to predict the execution time of transcoding tasks before their executions. Experiment results demonstrate the feasibility of cloud-based transcoding for live video streams and the efficacy of the proposed scheduling method in satisfying viewers' QoS demands without imposing extra cost to the stream provider.