Performance bounds and asymptotic optimality of modified (r, Q) policies for stochastic distribution inventory systems

We study a continuous-review, two-echelon inventory system with one central warehouse, multiple local facilities, and each facility facing random demand. Local facilities replenish their stock from the central warehouse (or distribution center), which in turn places orders at an outside supplier with an ample supply. Inventory replenishment at each location incurs a fixed-plus-variable cost for each shipment. The optimal policy remains unknown, and even if it exists, such a policy must be extremely complicated. Instead, we evaluate a class of easy-to-implement heuristics, called modified echelon (r, Q) policies. The parameters for such a heuristic are obtained by solving a set of independent single-stage systems. We show that the proposed policy is asymptotically optimal, as pairs of system primitives, such as the ratios of the fixed cost of the central facility to those of the local facilities, are scaled up. We also show that as the number of retailers grows, the performance bound of the heuristic converges to a primitive-dependent constant