The demand for global video has been burgeoning across industries. With theexpansion and improvement of video streaming services, cloud-based video isevolving into a necessary feature of any successful business for reachinginternal and external audiences. This paper considers video streaming overdistributed systems where the video segments are encoded using an erasure codefor better reliability thus being the first work to our best knowledge thatconsiders video streaming over erasure-coded distributed cloud systems. Thedownload time of each coded chunk of each video segment is characterized andordered statistics over the choice of the erasure-coded chunks is used toobtain the playback time of different video segments. Using the playback times,bounds on the moment generating function on the stall duration is used to boundthe mean stall duration. Moment generating function based bounds on the orderedstatistics are also used to bound the stall duration tail probability whichdetermines the probability that the stall time is greater than a pre-definednumber. These two metrics, mean stall duration and the stall duration tailprobability, are important quality of experience (QoE) measures for the endusers. Based on these metrics, we formulate an optimization problem to jointlyminimize the convex combination of both the QoE metrics averaged over allrequests over the placement and access of the video content. The non-convexproblem is solved using an efficient iterative algorithm. Numerical resultsshow significant improvement in QoE metrics for cloud-based video as comparedto the considered baselines.
展开▼