LaEC: Loss-aware Earliest Completion Data Scheduler for Multi-site Parallel Downloading

The multi-path data scheduler stands as the pivotal element profoundly influencing the performance of any multipath transport. Multi-site parallel downloading (MPD), which emerges as an alternative to the costly traditional dedicated Content Delivery Network (CDN), requests video segments from multiple economical edge data nodes simultaneously. However, the existing data scheduler for MPD solely prioritizes paths based on their minimal round trip time (RTT), which may lead to performance degradation in network environments with varying path latency and packet loss. To mitigate this issue, we present and implement LaEC (Loss-aware Earliest Completion), a novel data scheduler that considers both the latency and the path loss rate when selecting paths. We evaluated LaEC on the local test platform to assess its performance and the results demonstrate that LaEC outperforms the existing schedulers, LL and RR, with improvements of $5.59 \%-12.33 \%$ and $7.38 \%-15.57 \%$ respectively, in terms of download speed for video downloading tasks.