| With the appearance of new distributed multi-media applications and the full-speed development of network multi-media technology, highly efficient support of QoS will be more and more important. So many architectures have been proposed to support the quality of service. In those architectures, one of the most important problems is QoS routing, that is, to find the routes which satisfy the users' QoS demands.It has been proved that if there is more than two measures of the same kind, the QoS routing problem is a NP-complete problem~[8] So the QoS routing problem can not be solved by the shortest-path algorithms, let alone enumeration algorithm. To find the best path with multi-QoS constraints, a lot of heuristic algorithms have been proposed. But those researchs have different problems more or less. Especially they try to find the "best" path with unaccurate architectures and information. What's more, many researchs discuss the the QoS routing algorithms only, without considering the QoS architectures which are very important to the calculation of QoS measures.In this paper, we analyze and discuss on IP QoS architectures and packet scheduling algorithms, establish a more accurate mathematic model for QoS routing. We discuss the satisfaction optimization and many problems on it, and build up the satisfaction optimization model for QoS routing. What'more, we improve on the genetic algorithm, making it more suitable for QoS routing. On those bases, we try to solve the QoS routing problem. The emulational results indicate that the model can efficiently shorten the time of routing and enhance the success rate of QoS routing. What's more, it can avoid the congestion and make it possible for traffic symmetry.On those bases, we put forward a new multi-path routing algorithm for "best_effort" data, and discuss how to build up the routing table with satisfaction optimization and Genetic Algorithm. The emulational results prove that it is easy and effective to build up the routing table.
|
PDF Full Text Request