Capturing important statistics of a fading/shadowing channel for network performance analysis

We identify important characteristics of a fading/shadowing channel and present the work of measurement-based channel modeling for packet-level network queueing analysis. Our integration of wireless channel modeling and data queueing analysis at the packet-level provides a unique approach to study the effect of various channel dynamics on high-layer network performance, which otherwise cannot be captured through the traditional bit-level physical-layer channel modeling. In our study, the channel statistics are decomposed into three frequency regions [i.e., low (LF), mid (MF), and high (HF)]; the statistics in each frequency region is found to have significantly different impact on the queueing performance. While the HF statistics can be largely ignored in channel modeling due to their negligible impact on queueing performance, the LF statistics play the most important role in channel modeling because of substantial impact on queueing performance. Since the shadowing mainly represents the LF behavior of a channel, its dynamics are found to have a dominant effect on network performance as compared to the effect of multipath fading dynamics. In wireless networks, there are many other system factors which may change the channel dynamics, such as mobile user driving patterns, and forward-error-correction (FEC) coding (fixed or adaptive) using automated repeat request (ARQ) scheme. Our study further examines the individual impact of these factors on the network performance. In the measurement-based channel modeling, we use a Markov chain modeling technique to match the important channel statistics for queueing system analysis. The study shows an excellent agreement in queueing solutions between using the real original channel traces and using the sequences generated by the matched Markov chain models