Quantitative Study of an Outdoor Multi-hop 802.11 Network Performance using a Novel Passive Measurement Approach

This paper presents a novel passive measurement approach to accurately evaluate the performance of outdoor multi-hop 802.11 networks from the wireless side directly. Our approach employs five processing steps to evaluate per-hop system performances. Firstly, reference packets are identified from multiple independent packet traces. Secondly, multiple packet traces in one channel are merged together using the identified reference packets. Thirdly, redundant packets are filtered from the merging packet traces. To adjust for propagation differences between access points and monitors, the status of each packet is then corrected according to its context packets. Finally, an inference approach based on finite state machines (FSM) is designed to infer missing packets which are not recorded in any of the monitor traces but known to have existed in the real traffic. The unique features in our approach include propagation delay evaluation and FSM designs for existing packet status correction and missing packet inference. The measurement results collected from our outdoor multi-hop 802.11 testbed are used to validate the accuracy of our approach.