It is important to point out that our sole focus in this study is on the higher-level user request traffic to the Web site of interest. This is in contrast to the lower-level packet traffic that has been investigated in a considerable amount of previous work; eg, see [14, 20, 24, 19, 21, 22, 25, 5, 11, 12] and the references cited therein. While some previous studies have considered the user request traftic for certain types of Web servers (eg, see [1, 7, 2, 5, 13, 23, 15] and the references therein), these studies have not considered the user request traffic found at real, production-level, commercial Web sites. Moreover, to the best of our knowledge, our study is the first to demonstrate the important need to have the arrivals of user requests at a finer time scales than one second, and to develop a drill-down methodology to obtain an accurate representation of the user request interarrival process at such finer time scales. Several of the results of our analysis further provide differences with those of previous work, primarily reflecting the commercial Web site environments considered herein. Given the results of this analysis based on our drill-down methodology, we then model each Web server computing node as a general single-server queue with the arrival process based on the fractional Gaussian noise (FGN) model. We investigate the asymptotic behavior of the waiting times of user requests in this context. Previous work in this area has been mostly focused on either the performance impact of only the dependence structure while assuming deterministic service times, such as [17, 10], or the performance impact of only the heavy-tailed properties of the service times while assuming independent and identically distributed arrival processes [18]. In contrast, we address the performance asymptotics under both long-range dependent arrival processes and subexponential service time distributions. We develop lower bounds on the steady-state waiting time tall distribution asymptotics, which illuminate the different dominating components that influence server performance under various conditions.