Local Allocation of End-to-End Quality-of-Service in High-Speed Networks

Quality of service QOS requirements for applications in high speed networks are typically speci ed on an end to end basis Mapping this end to end requirement to nodal requirements facilitates providing QOS guarantees and simpli es connection admission In this paper we evaluate strategies for such local allocation of the end to end QOS A QOS allocation policy is said to perform better than another when the maximumnetwork load that it can support is greater A major contribution of this work is the development of a nodal metric that predicts the relative performance of QOS allocation policies in a network setting Computation of the nodal metric and direct evaluation of allocation policy performance for two simple network models yield valuable insight into the choice of allocation policies Intuitively one expects signi cant di erences in the performance of alloca tion policies when there are signi cant imbalances in nodal resource capacities or tra c loads It is found however that with the packet loss probability as the QOS metric there is little di erence in the performance of allocation policies in the regime of applications with low loss requirements From a practical viewpoint this suggests that a simple allocation policy may be adopted in this scenario with only a small decrease in carried load with respect to an optimal policy For applica tions which tolerate large packet loss or alternate QOS metrics however QOS allocation policies di er signi cantly in their performance Our results indicate that the development of optimal QOS allocation policies is of interest in such cases