MediaPaaS: A Cloud-Based Media Processing Platform for Elastic Live Broadcasting

摘要

Mobility is changing the way of how people consume live media content. By staying always connected with the Internet from various mobile devices, people expect to have enhanced TV viewing experience from anywhere on any device. Therefore, live broadcasting needs to be widely accessible and customizable, instead of being passive content only on TV. In this paper we present a cloud-based media processing platform, called MediaPaaS, for enabling elastic live broadcasting in the cloud. As an ecosystem-oriented solution for content providers, we outsource complex media processing from both content providers and terminal devices to the cloud. A distributed media processing model is proposed to enable dynamic pipeline composition and cross-pipeline task sharing in the cloud for flexible live content processing. Also, a prediction-based task scheduling algorithm is presented to minimize cloud resource usage without affecting quality of streams. The MediaPaaS platform allows third-party application developers to extend its capability to enable certain customization for running live channels. To our knowledge, this paper is the first work to openly discuss the detailed design issues of a cloud-based platform for elastic live broadcasting.