Throughput vs. Delay in Lossy Wireless Mesh Networks with Random Linear Network Coding

This work proposes a new protocol applying on-the-fly random linear network coding in wireless mesh networks. The protocol provides increased reliability, low delay, and high throughput to the upper layers, while being oblivious to their specific requirements. This seemingly conflicting goals are achieved by design, using an on-the-fly network coding strategy. Our protocol also exploits relay nodes to increase the overall performance of individual links. Since our protocol naturally masks random packet losses from the upper layers, it makes it particularly suitable for enhancing TCP's performance in wireless mesh networks, where packet losses are typically interpreted as a sign of congestion by TCP's congestion control algorithms, thus crippling TCP's throughput. To investigate the gains and downsides of our protocol, we implement it as a configurable proof-of-concept application, which is deployed and evaluated in a real test bed with Raspberry Pi devices. We show that order of magnitude gains in throughput over plain TCP are possible with moderate losses and up to two fold improvement in per packet delay in our results.