Topology information condensation in hierarchical networks

Abstract

Inspired by the PNNI protocol of the ATM Forum, this work focuses on the problem of node aggregation within peer groups and link aggregation between peer groups. It is assumed that the (large) network is already divided into peer groups. The objective is to maximally condense topology information subject to a given accuracy constraint. The QoS measures can be reduced into three distinct classes: an additive QoS class, a min-max one and the last, a combination of additive and min-max QoS measures. A new method for node aggregation of the additive QoS class (even with multiple QoS measures) is presented. The min-max class is discussed and Lee's optimal solution [Lee, W.C., Spanning tree method for link state aggregation in large communication networks, IEEE INFOCOM'95, 1995, pp. 297–302] for a single min(max) QoS measure is reviewed. Finally, we discuss the extension of our new method to the combination of additive and min-max QoS measures. A detailed example illustrates the presented algorithm for a single additive QoS measure. Subject to a given relative accuracy ε, it shows how to perform node and link aggregation on different hierarchical levels and how to establish the whole hierarchical structure of the original network.