A simulation study of pull systems with ascending/descending buffers and stochastic processing times

We study resource-balanced serial production systems with random processing times. We assume that the manufacturer controls buffer sizes and the mechanism in which material is released into the system. Our interest is in the impact different buffer arrangements have on work-in-process (WIP) and throughput (TP). In considering stochastically paced buffer-balanced systems, we find there is a ‘crossover point' at seven stages: for fewer than seven stages, it is best to select the system with the smallest buffers, while for more stages, it is best to select the system with the largest buffers. While the crossover point is robust relative to buffer sizes or processing time distribution, it is sensitive to the usage of ascending or descending buffers. We also find that as far as WIP and TP are the primary performance measures, systems with descending buffers should never be used.