New topology aggregation mechanisms for delay-bandwidth sensitive networks

Abstract

Today's networks are often hierarchically modeled for both scalability and security considerations. Topology aggregation is introduced for the purpose of simplifying routing by summering and compressing information of lower levels and advertising them to higher logical levels. However, the balance between performance and accuracy of state information must be taken into account since topology aggregation will ineluctably give rise to distortion. In this paper, new approximation methods based on approximation staircase formed by power-transformed linear regression model and hyperbolic-transformed linear regression model are proposed to approximate the service efficient frontier in the full mesh approach, and then the full mesh topology is transformed to a simpler tree structure and star structure. The proposed methods are shown to be valuable by performance analysis and comparison with existing methods.