Resource allocation for a stochastic CPM-type network through perturbation analysis

Abstract To minimize the expected length of a stochastic CPM-type network by allocating resources optimally is analytically insolvable and numerically impractical. The minimization is usually carried out by the steepest descent gradient search with simulation. Derivatives with respect to parameters of an allocation are estimated by simulating at different values of parameters. This article gives sufficient conditions under which derivatives of a class of these allocation problems can be estimated efficiently through simulating at one value of the parameters. An algorithm and an example are given to illustrate our procedure.