This thesis investigates two issues in operations research.

In the first part, we consider perishable inventory problems. Most early work ignores the clearance decision and focuses solely on the ordering decision until recently where heuristic clearance policies have been developed. In this paper, we approach the problem from a different angle by exploring its asymptotic behavior. Inspired by such asymptotic behavior, we examine simple policies that ignore clearance under minor conditions and establish theoretical bounds for them. The bounds not only vanish asymptotically, but also indicate a system size required to guarantee any given optimality gap. Numerical studies suggest that such policies can work very well for systems with reasonable sizes and practical management of complex pe...[ Read more ]

This thesis investigates two issues in operations research.

In the first part, we consider perishable inventory problems. Most early work ignores the clearance decision and focuses solely on the ordering decision until recently where heuristic clearance policies have been developed. In this paper, we approach the problem from a different angle by exploring its asymptotic behavior. Inspired by such asymptotic behavior, we examine simple policies that ignore clearance under minor conditions and establish theoretical bounds for them. The bounds not only vanish asymptotically, but also indicate a system size required to guarantee any given optimality gap. Numerical studies suggest that such policies can work very well for systems with reasonable sizes and practical management of complex perishable inventory systems is not so much harder than that of non-perishable ones.

In the second part, we consider a notion of “effective chaining" termed the Generalized Chaining Gap (GCG). Firstly we consider a production/distribution with customers arriving following a renewal process. This paper is among the first attempt to explore the benefit of flexibility for a continuous review system where the objective is to select the production capacity, demand fulfillment policy, and inventory control policy that maximize the expected infinite-horizon discounted profit. We show that, for any given connected network, a set of capacity decisions that satisfy the GCG is asymptotically optimal under continuous review. The second setting is the online demand fulfillment in a GCG network. Previous work shows that a network with positive Generalized Chaining Gap (GCG) is sufficient to achieve bounded performance and the performance bound is inversely proportional to the GCG. In this paper, we further extend to the case of (extremely) asymmetric demand structure and propose a dynamic load deviation policy that achieves bounded performance.