Reference queue tracking strategies for delay guarantees in ATM networks

A class of service scheduling schemes called the reference queue tracking (RQT) schemes is identified, and their ability to provide delay bounds on a per-VC basis is studied. A statistical multiplexer appears to an incoming VC stream as a single server queue whose service rate (bandwidth) is controlled by the service scheduling scheme. We refer to this queue as the virtual connection queue (VCQ). Corresponding to each VCQ, the RQT schemes use a fictitious single server reference queue (RQ) which is also fed with the VC stream but which is served at a constant rate equal to the allocated bandwidth of the VC. The RQT schemes serve VCs in such a way that some selected parameters of the VCQs track those of the corresponding RQs. An important property of a service scheduling scheme is the ability to guarantee that no cells depart from the VCQs later than from the corresponding RQs. Four schemes belonging to the RQT class are identified and their ability to satisfy the above property is investigated.