Routing Traffic with Quality-of-Service Guarantees in Integrated Services Networks

Transmission of interactive multimedia streams requires stringent Quality-of-Service (QoS) guarantees in delay, delay jitter, and bandwidth, which imposes strict resource constraints on the paths being used. While the general problem of finding a path that meets multiple QoS constraints is NP complete, we have shown that such a path can be found in polynomial time if the network service disciplines are rate proportional. However, two important issues remain unsolved. One issue is how to select among all feasible paths an efficient one that achieves high network throughput. We identify four optimality criteria and for each of them we propose a polynomial algorithm that selects paths that meet the criterion. Our simulation results show that, while the widest-shortest path performs best for heavy loads, the shortest-delay path has a performance edge for light loads. Surprisingly, the path that requires the reservation of the smallest amount of bandwidth performs the worst for both heavy and light loads. We also evaluate the influence of these routing algorithms on the throughput of best effort traffic. Our results show that, using the shortestwidest path algorithm for routing guaranteed traffic can result in relatively low throughput for best-effort traffic, while the other three algorithms result in comparable performance. The second open issue is that, although these algorithms are polynomial, they have much higher computational complexity than the Bellman-Ford algorithms used in today’s Internet. We propose approximation algorithms that have a running time within a constant factor of that of the Bellman-Ford algorithm without sacrificing network throughput. Overall, the approximation algorithms that select the shortest-delay path results in consistently high network throughput.