SERVER STAFFING TO MEET TIME-VARYING DEMAND

We consider a multiserver service system with general nonstationary ar rival and service-time processes in which s(t), the number of servers as a function of time, needs to be selected to meet projected loads. W e try to choose s(t) so that the probability of a delay (before beginn ing service) hits or falls just below a target probability at all time s. We develop an approximate procedure based on a time-dependent norma l distribution, where the mean and variance are determined by infinite -sewer approximations. We demonstrate that this approximation is effec tive by making comparisons with the exact numerical solution of the Ma rkovian M(t)/M/S-t model.