Title: Modeling of Web User Behavior and User Perceived Website Availability
Abstract: Depending upon the user behavior user perceived availability may differ from the actual system availability. Xei et al [5] studied the relationship between the system availability and the user-perceived availability and their distribution with change in system MTTF (Mean Time To Failure) and MTTR (Mean Time To Repair) using 3 different mathematical models of user-perceived unavailability, called SUSH, SUMH and DPF models respectively as shown. We corrected SUSH model and extended DPF model to bring it closer to the real world. For lower repair rate, CTMC analysis underestimates perceived unavailability and at higer repair rate it overestimates. We verified our claims from data obtained for user-server interactions with real web servers. We have obtained encouraging results, however, we need to go for more extensive data collection before drawing any conclusions. We modified DPF model to bring it more closer to real world by incorporating generally distributed transitions. This is a very complex model and not amenable to numerical solutions in its given form. So, we outlined solution for a simpler model leaving more extensive models analysis for future work.
Thesis: [pdf]