Linear Approximations for Stochastic Serial Network Performance

Many systems are characterized by the flow of entities through a network of servers. Systems may have random, non-stationary interarrival times and arrival group sizes, stochastic service times, parallel servers, random routing between service nodes, and finite buffers that induce blocking. Representative systems include the flow of batches or materials through a manufacturing system, the processing of paperwork through a service system, or passengers passing through a security checkpoint. In all cases, the flow is subject to conservation equations that can be used to develop deterministic approximations of actual throughput rates, queue lengths, and throughput times associated with a particular network architecture and parameterization. This project aims to investigate the precision of deterministic flow-based approximations to estimate the performance metrics of such networks.