A repair time model of a web based system including administrator working hours

The paper presents a web based system reliability analysis. We propose to model different types of faults (hardware one, software and security incidents) taking consideration only the effect of the failure, not the source of the fault. It is assumed that failure events are independent and the time to failure is exponential. Whereas the time to repair is not exponential since repair actions are taken only when administrators are at work. We assume that administrators are not working 24/7. The paper presents an algorithmic model of the repair time including administrator working hours. The model is used to estimate reliability parameters (mean time, standard deviation and 90th percentile of yearly down time) by a use of Monte-Carlo simulation. The numerical results are compared with results from the analytical model (Markov one) that assumes exponential distribution of all repair times. Results for two web exemplar web systems (with reliability model consisting of two and five states) show the range of error caused by the exponential distribution assumption.