Dispatching and Allocating Servers to Stochastically Failing Networks

The problem considered is that of allocating and dispatching servers to networks whose nodes and/or arcs fail stochastically. A mathematical programming formulation is developed which allocates servers to nodal demands so that expected loss is minimized. From this a zero-one programming problem is formulated to insure whole server allocation and to allow for decreasing service time and cost functions in the number of servers dispatched to a demand. Comparative computer experience with two zero-one codes is reported. Server transfers between subparts of networks, use of the model for prepositioning servers throughout the network and other extensions are discussed in the final section.