Inter-Stream Adaptation over Group Reservations

In this paper, we address the issue of dynamically allocating resources to a group of streams belonging to a distributed multimedia application that reserves a collective amount of resources to be shared by its streams. The traac characteristics of the streams are speciied using the M-LBAP (Modiied Linear Bounded Arrival Processes) model. The M-LBAP model provides tight characterization for the traac while maintaining the simplicity and linearity of the LBAP model. Delay bounds for streams sharing a group reservation are analytically derived using the M-LBAP model. Degradation paths speciied using the M-LBAP model are used as the basis for inter-stream adaptation in order to dynamically allocate the reserved resources among the group of streams belonging to the application. A dynamic rate based approach for inter-stream adaptation (RISA) is introduced and contrasted to static resource allocation policies. It is shown that higher utilization and acceptance ratios are achievable by RISA over the static policies. These achievable results are reeected on and summarized by a proposed metric for judging the the eeectiveness of resource allocation on the overall quality of session.