Reducing traffic fluctuations of link state QoS routing algorithms in virtual circuit networks

Abstract

Quality of Service (QoS) routing methods are expected to replace existing intradomain routing protocols in future QoS-based data networks. QoS routing allows the selection of feasible paths for connections requiring QoS support. Link state QoS routing performs path selection based on traffic parameters carried within the requests and information about available network resources provided by link state advertisements. A good link state QoS routing method should efficiently utilise networking resources and should incur little overhead in spreading information about resource availability.When a network implements load sensitive routing, traffic fluctuations are often observed which give rise to routing instability. This results in poor route selection and an overall degradation of network performance. Such oscillatory behaviour may be expected in datagram networks, because of their rapid response to routing updates, but it is not peculiar to them. It also affects virtual circuit networks, when the update interval is comparable with the average duration of a connection.In this paper we present methods for reducing this effect in virtual circuit networks. The first method--ALCFRA (Adaptive Link Cost Function Routing Algorithm) can be used in networks with sparse connectivity, while the second algorithm--CAR (Connectivity Aware Routing) is designed to work well in the majority of network topologies. Such methods improve the routing stability, which in turn leads to a higher acceptance ratio of arriving connections. The route selection improvements presented in this paper not only stabilise the link state QoS routing methods, but also achieve excellent performance over a variety of network topologies.