Abstract

Using detailed statistical analysis of data taken from simulations and test-bed experiments, we have recently reported the inappropriateness of a queue-decoupling approximation that has been implicitly adopted by many distinct authors to incorporate queueing behavior in analytic models of finite load 802.11 networks. In the present paper we show that this flawed hypothesis leads to network throughput prediction errors in the presence of stations with asymmetric offered loads. We suggest that care should be taking in drawing deductions from models that adopt this hypothesis and that further modeling innovation is necessary.