Localized Route Selection for Traffic with Bandwidth Guarantees

Abstract

Localized Quality of Service (QoS) routing has recently been proposed as a viable alternative approach to traditional QoS routing algorithms that use global state information. In this approach, problems associated with maintaining global state information are avoided by having the source nodes to infer the network QoS state based on flow blocking statistics collected locally, and perform flow routing using this localized view of the network QoS state. One of the most important QoS metrics is bandwidth and in this paper we introduce a localized routing scheme for providing bandwidth guarantees, based on a crediting scheme that uses an estimate of flow blocking probability to adjust credits and select routes. We compare its performance against the localized proportional sticky routing (PSR) algorithm using different types of network topologies, QoS requirements and traffic patterns and under a wide range of traffic loads. Extensive simulations show that our algorithm outperforms the PSR scheme while maintaining the same level of adaptivity and stability. We also compare performance against the well known widest-shortest path (WSP) algorithm that uses the global network QoS state and show that our algorithm gives a comparable performance with better time complexity and very much lower communication overhead. The simulations were implemented using OMNET ++, a discrete-event simulator that provides a rich set of functions and tools for simulating the elements of a communication network, such as nodes, links and packets.