Finite population task oriented queuing: a model for Internet server capacity planning

A new model is developed and solved for Internet server capacity planning. This model includes details that have not been done in one model. These details include: finite population; general service times, and task oriented queuing with the first task being of a higher priority. In developing this model a new way to derive the M/G/1/K/K general service time machine repairman model is developed. It has the advantage of being more intuitive to derive compared with previous derivations. It has the additional advantage of being recursive on population size. All the analytic model results give good agreement with the discrete event simulation results.