Optimal buffer allocation in a multicell system

Motivated by a layout design problem in the electronics industry, we study in this article the allocation of buffer space among a set of cells. Each cell processes a given part family and has its own revenue-cost structure. The objective of the optimal allocation is to maximize the net profit function (total production profits minus total buffer allocation costs). According to the flow pattern of jobs, the cells are categorized into two types. A type 1 cell is modeled as a Jackson network; a type 2 cell is modeled as an ordered-entry system with heterogeneous servers. Both models have finite waiting room, due to the buffer capacity allocated to the cells. We show that under quite general conditions, the production rate of each cell of either type is an incresing and concave function of its buffer allocation. Exploiting this property, a marginal allocation scheme efficiently solves the optimal buffer allocation problem under increasing concave production profits and convex buffer space costs.