Delay guarantee and bandwidth allocation for network services

Abstract

This paper presents a packet scheduling scheme for ensuring delay and bandwidth as a quality of service (QoS) requirement. For customers, rightful service is given while optimizing revenue of the network service provider. A gradient and fixed point type algorithms for updating the weights of a packet scheduler are derived from a revenue-based optimization problem. In the linear pricing scenario, algorithms are simple to implement. We compared algorithms with optimal brute-force method. Especially fixed point algorithm converges very fast to the optimal solution, typically in one iteration and about 40 operations, when number of classes is three. The weight updating procedures are independent on the assumption of the connections' statistical behavior, and therefore they are robust against erroneous estimates of statistics. Also, a call admission control (CAC) is implemented in context of our scenario.