Traffic models and admission control for variable bit rate continuous media transmission with deterministic service

Future high-speed computer networks are projected to carry a dizzying array of stored continuous media (CM) flows emanating from sources such as digital libraries and video servers. These flows are often variable bit rate in nature and are characterized by tight delay requirements. In order to guarantee such delay requirements, the network must reserve resources at the links of the path for the given CM flow. The development of an efficient algorithm for deciding whether a new flow can be supported while guaranteeing delay requirements for it and all existing flows is a challenging problem which has been the subject of considerable research [3, 6, 5, 11, 12, 13]. Much of the research in recent years has focussed on resource reservation and call admission for the case when hard guarantees are required. In this case much of the attention has focussed on the development of algorithms that take time-invariant descriptions of CM flows as inputs [5, 12, 13]. A CM flow description is said to be time-invariant if the description over an interval of any arbitrary length does not depend on the starting point of the interval. The focus of much of these works has been on the development of parsimonious time invariant flow descriptions that give rise to efficient call admission algorithms that provide high resource utilization. Unfortunately these algorithms are unable to generate high resource utilization when offered variable bit rate flows with stringent delay requirements (100500ms). In this paper we take a different approach, namely, we address the problem of developing call admission algorithms that use more general descriptions of stored CM flows which are not necessarily time-invariant. We begin by demonstrating that the use of a complete description of a stored CM flow (e.g., frame sizes in the case of video) can result in a substantial increase in the number of supported flows over what is possible using time-invariant flow descriptions. Since the computational requirements of the algorithm are proportional to the length of the CM flow (size of the video), we present an algorithm for producing a parsimonious flow description which still results in efficient call admission. We show that using this simple algorithm, the number of supported flows is substantially larger than the number of flows that can be supported using time-invariant flow descriptions.