Flexible QoS Route Selection with Diverse Objectives and Constraints

Many applications need routing paths with diverse Quality of Service (QoS) objectives and constraints based on multiple metrics, including delay, bandwidth, and error rate. Since most multi-metric path selection problems are NP complete heuristics are used. However, current heuristics tend to be specific to a problem, computationally complex, or amalgamate metrics into one cost (which limit the allowable solution space). This paper investigates: (a) an approximation that can quickly find a ´good´ path subject to flexible objectives and constraints, without amalgamating link metrics, and (b) the application of simulated annealing to improve path quality. Results for networks of 100 to 1000 nodes show the new approximation reduces min-max path cost by 28% over existing multi-metric heuristics (and by 40% over single metric heuristics); simulated annealing, however, could only provide a further 3% improvement.