Predicting the performance of ad hoc networking protocols for mesh networks has typically been performed by making use of software based simulation tools. Experimental study and validation of such predictions is a vital to obtaining more realistic results, but may not be possible under the constrained environment of network simulators. This paper presents an experimental comparison of OLSR using the standard hysteresis routing metric and the ETX metric in a 7 by 7 grid of closely spaced Wi-Fi nodes to obtain more realistic
results. The wireless grid is first modelled to extract its ability to emulate a real world multi-hop ad hoc network. This is followed by a detailed analysis of OLSR in terms of hop count, routing traffic overhead, throughput, delay, packet loss and route flapping in the wireless grid using the hysteresis and ETX routing metric. It was discovered that the ETX metric which has been extensively used in mesh networks around the world is fundamentally flawed when estimating optimal routes in real mesh networks and that the less sophisticated hysteresis metric shows better performance in large dense mesh networks.