A comparison of cross- and single-layer retransmission optimisations for 802.11

Retransmissions in the 802.11 wireless LAN protocol often contain data that was successfully conveyed in the initial message, but is ignored due to corruption during transmission, thereby reducing network efficiency. By compartmentalising data, and consequently the corrupted portions of the frame, into distinct sub-fragments 802.11 protocol, performance can be further improved. Cross-layer designs provide opportunities to assist in accomplishing this improvement that may not otherwise present themselves in normal architectures using isolated layers. This thesis presents two new and novel schemes, one that is based on cross layer design and the other using the traditional single layer approach. Both schemes are based upon the same principle of compartmentalising corruption. Additionally, a comparison between the two schemes is provided in order to investigate the strengths and weaknesses of cross-layer design, while simultaneously presenting an 802.11 protocol with enhanced spectral effciency. Both schemes divide layer 2 frames into small fragments separated by frame check sequences in order to facilitate retransmission of data that was corrupted during transmission, while avoiding resending correctly received data. They differ in the metadata mechanisms relied upon to manage this process, and the layer in which retransmission decisions are made, but are fundamentally similar; permitting a unique opportunity for the informative comparison between cross- and single- layer designs. Moderate bandwidth increases were observed over standard 802.11n with both schemes at medium to high channel error rates, alongside situational decreases in latency. The cross-layer scheme performed slightly worse than the single-layer scheme, however it provided a significant opportunity for further optimisations that were not readily apparent in the single-layer scheme.