Performance bounds for scheduling queueing networks (Q1198582)

Some general classes of multi-customer networks are compared with (partially very) simplified derivates. Roughly the idea is: Forget about those structures of the system which makes analysis complex and try to do this in a way such that the new model when run in parallel to the old one can be strictly compared either by pathwise comparison of the describing processes or by comparing the steady-state expectations of the performance measure. Some of the bounds and concerned numerical results are called ``moderately encouraging'' by the authors.