Effective rate per user on shared broadband wireless data channels

If N users share a wireless channel of capacity R on a time-division basis, the average rate is R/N, but this is not necessarily the apparent or effective rate experienced by the user. If the user is engaged in activities that demand bursts of data followed by inactive periods, such as downloading and reading Web pages, then the channel rate as perceived by the user depends on how long it takes to load the web page following the user's request. This paper develops simple queueing models and a simulation to show how this effective rate depends on R and N, and the parameters of the data traffic. The purpose here is not to provide a rigorous or exhaustive analysis, but rather to develop an initial understanding of the relationship among channel rate, number of active users, idle time, data block size, and the effective rate. These relationships are useful for wireless link design and tradeoff analysis