A bandwidth allocation model for a two-pass RSVP setup mechanism

Abstract

Traditional one-pass RSVP may encounter a so-called killer reservation problem (KRP), when more than two reservation requests from different receivers are merged. Two-pass with advertising (TPWA) mechanism [RSVP Killer Reservation, Internet Draft, IETF (1999)] is, therefore, presented to inherently avoid the KRP. In this paper, we present an analytical model to reasonably allocate bandwidth in TPWA. TPWA issues pre-engaged bandwidth request at the first pass before effectively reserving bandwidth at the second pass. Since the pre-engaged bandwidth may be used to deliver non-QoS traffic before the second pass is actually enforced, allocation of adequate amount of pre-engaged bandwidth at a router becomes an important issue. We build a model to analyze the resource allocation in TPWA based on recursive law and Markov chains. In the model, we are interested in determining an optimal resource distribution among the pre-engaged bandwidth, the free bandwidth and the reserved bandwidth under the influence of the RSVP parameters, including message request rates, session lifetime, link-state refresh periods, timeout factor, and message loss probability.