ABSTRACT
Some recent results are surveyed and unified showing that bias terms of Markov reward structures are essential to conclude error bound or monotonicity results for steady-state measures when dealing with
perturbations
finite truncations
system modifications, or
system comparisons (bounds)
The results are illustrated by a non-product-form queuing network example with practical phenomena such as blocking, overflowing, and breakdowns. Monotonicity results and explicit error bounds are hereby obtained.
