Browse Theses

The burstiness of variable bit rate (VBR) traffic makes it difficult to efficiently utilize network resources, as well as to provide guaranteed end-to-end network quality of service (QoS) to the traffic sources. Smoothing or shaping the traffic at the entrance of the network reduces the burstiness thus allowing for higher utilization within the network since less network resource is required for the smoothed traffic. On the other hand, quality of service requirements for network clients are typically end-to-end requirements, which impose corresponding performance requirements on both the network and the end-systems. In this thesis, traffic shaping algorithms that satisfy a given set of delay and buffer constraints for isochronous VBR traffic are introduced. The novel features include a unique solution for both stored and real-time traffic; the ability to optimize algorithm behavior for a specific network service; and robustness to traffic estimation error. The effect of traffic shaping on end-to-end quality of service is investigated for packet-switching networks with deterministic performance guarantees and it is proved that by smoothing VBR traffic, more connections can be supported for the same QoS when the network is congested.

A new bandwidth renegotiation algorithm is developed by integrating traffic shaping with bandwidth allocation. This novel approach allows for tracking the bandwidth requirements of the traffic using a moderate renegotiation rate which results in higher bandwidth efficiency. The features include the provision of universal interoperability by decoupling the source from the network, deterministic delay bounds, and constant-quality video transmission with negligible cell-tag rate.

Finally, a novel concept called aggregate smoothing is introduced which integrates multiplexing of VBR traffic with traffic shaping. It is demonstrated that the number of rate changes and variation of the combined traffic are significantly reduced allowing for cost-effective transport of real-time traffic. This result is particularly important for public networks carrying aggregated traffic since it is shown that aggregate smoothing increases the network utilization.