1. QoE-aware distributed cloud-based live streaming of multisourced multiview videos
- Author
-
Aiman Erbad, Kashif Bilal, and Mohamed Hefeeda
- Subjects
Crowdsourced live video ,Multimedia ,Computer Networks and Communications ,business.industry ,Video capture ,Computer science ,ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION ,020206 networking & telecommunications ,Cloud computing ,02 engineering and technology ,Transcoding ,computer.software_genre ,Video quality ,Live streaming ,Computer Science Applications ,Multiview video streaming ,Hardware and Architecture ,0202 electrical engineering, electronic engineering, information engineering ,020201 artificial intelligence & image processing ,The Internet ,Quality of experience ,Resource allocation ,business ,computer - Abstract
Video streaming is one of the most prevailing and bandwidth consuming Internet applications today. Advancements in technology and prevalence of video capturing devices result in massive multi-sourced (aka crowdsourced) live video broadcasting over the Internet. A single scene may be captured by multiple spectators from different angles (views), enabling an opportunity for interactive multiview video by integrating these individually captured views. Such multi-sourced multiview video offers more realistic and immersive experience of a scene. In this paper, we present a Quality of Experience (QoE) driven, cost effective Crowdsourced Multiview Live Streaming (CMLS) system. The CMLS aims to minimize the overall system cost by selecting optimal cloud site for video transcoding and the number of representations, based on the view popularity and viewer's available bandwidth. In addition, we present a QoE metric considering delay and received video quality. We formulate the selection of optimal cloud site and number of representations to meet the required QoE as a resource allocation problem using Integer Programming (IP). Moreover, we present a Greedy Minimal Cost (GMC) algorithm to perform resource allocation efficiently. We use real live video traces collected from three large-scale live video providers (Twitch.tv, YouTube Live, and YouNow) to evaluate our proposed strategy. We evaluate the GMC algorithm considering the overall cost, QoE, video quality, and average latency between viewers and transcoding location. We compare our results with the optimal solution and the state-of-the art policy used in a popular video steaming system. Our results demonstrate that the GMC achieves near optimal results and substantially outperforms the state-of-the art policy. This publication was made possible by NPRP grant # [ 8-519-1-108 ] from the Qatar National Research Fund (a member of Qatar Foundation). We are thankful to the Denny Stohr for providing YouNow dataset. The findings achieved herein are solely the responsibility of the author[s]. Scopus
- Published
- 2018
- Full Text
- View/download PDF