Throughput upper bounds for open Markovian queueing networks with blocking

We present a simple but effective decomposition method which yields tight throughput upper bounds for open Markovian finite-buffered queueing networks under any commonly used blocking mechanism. This method is first elaborated on with three specialized network configurations, tandem, split and merge, and then extended to a general configuration. Also shown is that the existing throughput-bounding methods for open queueing networks with blocking can be improved via duality consideration. Computational experiments confirm that our method is superior to all other existing methods.