Towards efficient and scalable implementation for coding-based on-demand data broadcast

Abstract

Network coding has been demonstrated as a promising solution to further enhancing the bandwidth efficiency for on-demand broadcast. In this work, first, we show the performance improvement of a straightforward implementation of coding based on-demand data broadcast algorithms over the traditional on-demand broadcast approaches. Second, as the straightforward implementation of the optimal approach has overwhelming computational overhead, we propose an efficient generalized implementation scheme, which can be applied to all the existing on-demand scheduling algorithms. The proposed scheme reduces the computational overhead while achieves the same performance as the straightforward implementation. Third, to further enhance system scalability, we propose an approximate implementation method with even lower computational overhead while maintaining near optimal performance. Finally, we conduct an extensive simulation study and the results demonstrate that the proposed efficient implementation schemes can improve the system performance over 40% compared with the traditional broadcast approach, and the computational overhead can be reduced by 75% compared with the straightforward implementation. In addition, we show that the proposed approximate implementation can further reduce the computational overhead significantly and it is able to strike a balance between the service performance and system scalability.