CISUC

Quality of Service Routing for Class-Based Networks

Authors

Abstract

The communication through the Internet has been changing due to the emergence of new types of applications. Nowadays, applications such as video-conference, IP telephony and distributed games are becoming a reality for a large group of users. In this context, the best-effort service traditionally provided on the Internet is insufficient for the support of such a diversity of applications.
There have been several proposals to overcome the limitations of the best-effort service, in order to provide Quality of Service (QoS) for the various new types of applications. Nevertheless, although some networks already have mechanisms for the support of Quality of Service, Quality of Service routing is still an open issue due to the level of complexity of the problem. The identification of the open issues in the Quality of Service routing domain was at the basis of this work presented this thesis.
The main approaches to the Quality of Service routing problem are mostly aimed at service models where there are performance guarantees, associated with resource reservation and with routing at the flow level. This work addresses the problem of Quality of Service routing in networks where traffic differentiation is class-based and where strict guarantees are not supported. Therefore, it is proposed a Quality of Service routing strategy, named LCT-QoSR (Laboratory of Communications and Telematics-Quality of Service Routing).
The QoS routing problems analysed in this thesis are the following: the algorithmic complexity associated with the computation of paths that satisfy a set of Quality of Service characteristics; the overhead introduced in the network elements and in the communication system; the inaccuracy in routing information used for Quality of Service aware path selection; the instability associated with routing oscillations.
The complexity of QoS routing is due to the need of performing path computation subject to multiple constraints. The main algorithms for the solution of this problem and the approach followed on the proposed routing strategy are presented in this thesis.
The overhead associated with QoS routing has two main sources, namely, the routing messages needed to keep the routers with an up-to-date view of the state of the network and the processing effort for the constraint-based path computation. The main solutions to these problems and a set of mechanisms to overcome them within the LCT-QoSR strategy are presented in this thesis.
The main sources of routing information inaccuracy are identified, specifically, routing information aggregation in hierarchically organized networks and stale routing information, among others. Moreover, the most relevant approaches that address this problem are presented and the behaviour of the LCT-QoSR strategy concerning this issue is evaluated.
The instability problem due to routing oscillations is discussed in this thesis, and a set of approaches to avoid it are presented. The relevance of the instability problem in the context of this work led to the proposal of the Class-Pinning Election for Loop Free Quality of Service Routing, an extension to the basic LCT-QoSR routing strategy proposed.
The results of the evaluation of the Quality of Service routing strategy proposed are presented and discussed in the thesis, at the light of the above mentioned Quality of Service routing problems. Furthermore, the contribution of the proposed strategy for traffic performance and for the support of different types of applications is assessed.
The results obtained showed that the mechanisms proposed in the LCT-QoSR strategy satisfy the objectives defined, by contributing to an increased traffic performance and better resource utilization, however, without introducing an excessive overhead in the network.

Keywords

QoS routing, Differentiated Services, class-based routing

Subject

QoS routing

Related Project

IST FP6 NoE E-NEXT: Emerging Networking Experiments and Technologies

PhD Thesis

Quality of Service Routing for Class-Based Networks, October 2005

Cited by

No citations found