Probabilistic chunk scheduling approach in parallel multiple-server DASH

Li Liu,Chao Zhou,Xinggong Zhang,Zongming Guo,Cheng Li
DOI: https://doi.org/10.1109/VCIP.2014.7051490
2014-01-01
Abstract:Recently parallel Dynamic Adaptive Streaming over HTTP (DASH) has emerged as a promising way to supply higher bandwidth, connection diversity and reliability. However, it is still a big challenge to download chunks sequentially in parallel DASH due to heterogeneous and time-varying bandwidth of multiple servers. In this paper, we propose a novel probabilistic chunk scheduling approach considering time-varying bandwidth. Video chunks are scheduled to the servers which consume the least time while with the highest probability to complete downloading before the deadline. The proposed approach is formulated as a constrained optimization problem with the objective to minimize the total downloading time. Using the probabilistic model of time-varying bandwidth, we first estimate the probability of successful downloading chunks before the playback deadline. Then we estimate the download time of chunks. A near-optimal solution algorithm is designed which schedules chunks to the servers with minimal downloading time while the completion probability is under the constraint. Compared with the existing schemes, the experimental results demonstrate that our proposed scheme greatly increases the number of chunks that are received orderly.
What problem does this paper attempt to address?