Discrete Water Filling Multi-Path Packet Scheduling

Abstract

We study the performance of a coded point-to-point multi-path (MP) packet erasure channel (PEC) network model consisting of one sender (Tx) and one receiver (Rx). A network coded discrete water filling (DWF) scheduler is the core invention of this work. We provide an optimization framework to allocate coded packets over multiple network paths of varying channel conditions while minimizing the transmission delay. Applying the DWF framework to a feedback-based protocol shows significant throughput gains, delay and efficiency improvements compared to single path (SP) systems: In an example network with 4 paths we improve the transmission rate by a factor up to 2. This is not only beneficial for throughput-demanding applications such as large file downloads, but also for real-time systems such as livevideo streams which require low-latency environments. Moreover, we provide the optimization formulation for the DWF algorithm and a low-complexity implementation. The presented findings pave the way for efficient scheduling in next-generation transmission protocols for network coded MP and mesh networks.