Abstract
This dissertation studies the trade-off between route adaptation and persistence. Our thesis is that adding route persistence to shortest-path routing can enhance network performance, especially under heavy traffic conditions. Shortest-path routing can cause route oscillation and instability, thereby increasing congestion and reducing the effective throughput of the network.
To study the effect of route persistence on network performance, the dissertation inroduces a new class of routing techniques, called semi-persistent techniques, that offer a trade-off between route adaptation and persistence. Semi-persistent techniques add route persistence to shortest-path computation. This route persistence reduces oscillation by reducing the number of routes that shift from high-traffic links to low-traffic links. With various levels of route persistence, semi-persistent techniques exhibit multiple routing behaviors that span a spectrum between shortest-path and static routing.
This dissertation offers a promising advance for network routing. Simulation results show that certain semi-persistent techniques achieve significant throughput increases over shortest-path routing for a majority of the studied topologies and traffic loads. With further study of the relationship between traffic load and the persistence level of maximal throughput, semi-persistent techniques may be designed to effectively adjust their routing behavior to suit current network conditions.
