The method of surrogate delays: Simultaneous resource possession in analytic models of computer systems

This paper presents a new approach to modelling the simultaneous or overlapped possession of resources in queueing networks. The key concept is that of iteration between two models, each of which includes an explicit representation of one of the simultaneously held resources and a delay server (an infinite server, with service time but no queueing) acting as a surrogate for queueing delay due to congestion at the other simultaneously held resource. Because of this, we refer to our approximation technique as the “method of surrogate delays”.