Generalized scheme for topology-update in dynamic networks

Abstract

Many network control-functions like routing and flow-control require the knowledge of the whole network topology at each network site. The topology-update problem is the problem of keeping this knowledge current, when the network topology is dynamic, i.e. when links fail and recover at arbitrary times. Existing networks cope with the problem by the use of flooding algorithm, indicating the most recent topology status either by employing sequence numbers, which are akin to time-stamps, or by spontaneous periodic flooding of the most recent information. In contrast, recently Spinelli provided a topology-update algorithm which does not use sequence numbers and is event-driven.

Spinelli's algorithm has unbounded message complexity. This paper generalizes and modifies Spinelli's algorithm to a class of algorithms of bounded message complexity, thus making the algorithm practical. Moreover, the generalization provides a spectrum of algorithms out of which one can choose the one which best suits the problem at hand.

Preliminary Report

Supported by NSF Presidential Young Investigator Award under grant DCR84-51396 & matching funds from IBM Faculty Development Award under grant D840622.