Asymptotically optimal interruptible service policies for scheduling jobs in a diffusion regime with nondegenerate slowdown (Q415638)

The goal is to study a control problem with a single pool of identical servers, some customer classes in case of heavy traffic. The service time is exponential and the cost is associated with headcount and queue-length processes, rather than delay. \({\hat X}_n\) and \({\hat Q}_n\) denote the diffusion scale deviation of the headcount process from the quantity corresponding to the underlying fluid model and the diffusion scale queue-length. The authors consider minimizing \[ c_X^n=\int_0^uC({\hat X}_n(t))dt\quad \text{and}\quad c_Q^n=\int_0^u(C({\hat Q}_n(t))dt, \] where \(C\) is a continuous function, nondecreasing in the usual partial order, \(u>0\) is a constant. It is shown that any weak limit point of \(c_X^n\) stochastically dominates \[ \int_0^uC^*(W(t))dt \] for a suitable one-dimensional reflected Brownian motion \(W\), there is a policy which asymptotically achieves this lower bound. For \(c_Q^n\) a similar result is proved.