Bounded-cell-loss-ratio flow control and reliable ABR multicast

In this dissertation, we focus on the development for Available-Variable-Bit (ABR) flow control in ATM networks. We give a sufficient condition of max-min fairness convergence for ER-based ABR flow control. The tractability of this sufficient condition is verified by deriving three switch algorithms.; Bounding the loss ratio is another critical issue in flow control design. We use a learning model called Hedge Boosting and recursive least-squares estimation to capture the long-range dependence. With the on-line prediction, the cell loss ratio is bounded below a given ratio by reducing the available bandwidth with a number calculated under the Gaussion process assumption.; We study how to extend error recovery for ABR multicast (one-to-many communication). We use the backward resource management cells for ABR flow control to carry error information, and design an error control algorithm for multicast. In most of time, the algorithm will forward the retransmission cells only to the destinations requesting those cells.; At last, we prove that with max-min fairness guarantee, for a multicast session with two link-disjoint connections to support fault tolerance, a third connection atop these two existing connections could be available for the multicast session.