Summary: It is very important in many real-life systems to decide when the server should start his service because frequent setups inevitably make the operating cost too high. Furthermore, today’s systems are too intelligent for the input to be assumed as a simple homogeneous Poisson process. In this paper, an M/G/1 queue with general server setup time under a control policy is studied. We consider the case when the arrival rate varies according to the server’s status: idle, setup and busy states. We derive the distribution function of the steady-state queue length, as well as the Laplace-Stieltjes transform of waiting time. For this model, the optimal \(N\)-value from which the server starts his setup is found by minimizing the total operation cost of the system. We finally investigate the behavior of system operation cost and the optimal \(N\) for various arrival rates by a numerical study.