An increasing number of real-time applications have put new demands on the Internet's traditional routing protocols and mechanisms that provide only best-effort service. To provide the applications with quality of service (QoS), a paradigm of the QoS-based routing infrastructure, based on newly proposed routing protocols, service models, signaling mechanisms, may be defined. The goal of this research aims to design new mechanisms and schemes to reduce the overhead, and achieve good scalability of the underlying paradigm of the QoS-based routing without sacrificing its performance. Three are there parts contained in this research. First, we address two issues, which are maintaining the certain level of the knowledge of the network resource availability and achieving the scalable routing information advertisement in implementing the intradomain QoS-based routing. We test three advertisement schemes to advertise the available bandwidth. We provide some heuristic suggestions on achieving the good performance and maintaining the scalability of the QoS-based routing information advertisement. By employing the proposed crankback with cache, the performance can be boosted, while using large values of a clamp down timer to reduce the advertisement traffic. Second, we have designed aggregation algorithms for asymmetric QoS-based routing information (e.g., link residual bandwidth), using the mesh, the star, and the spanning tree of the existing aggregation approaches. With a detailed comparison, we have shown that the spanning-tree approach with the min-tree approximation to handle the link asymmetry is the most promising considering the trade-off between its performance and the amount of advertisement traffic that is generated. Third, we propose Maximal Reservable Bandwidth Tree (MRBT) algorithm that is a path-finding algorithm for the QoS-based routing, and aims to aggregate flows traveling thorough a transit routing domain to further reduce the amount of the stored state information and improve the scalability of the per-flow traffic management. We explain the construction of MRBTs in detail and their bandwidth update procedures, which are “Establish” and “Release.” We demonstrate, through extensive simulation, that MRBT could achieve good scalability and maintain the performance, compared to the conventional Dijkstra's path-finding algorithm for QoS-based routing.
展开▼