Backup VP reservation method using multiagents and its experimental evaluation

Abstract

AbstractHigh reliability schemes of switching from a primary VP to a backup VP by establishing two virtual paths (VPs), a primary VP and a backup VP, between a source (S) node and a destination (D) node in an ATM network when a failure of the node or the link occurs on the primary VP are known. The schemes performing the switching of the VPs without using multiagents have a problem associated with the incapability of switching rapidly from a primary VP to a backup VP, with the increase in the scale of the networks and the number of primary VPs (and backup VPs). In Ref. 2, a scheme of establishing backup VPs using multiagents such that such switching is performed rapidly is proposed. In this scheme, under the assumption that a candidate backup VP is given for each pair of an S node and a D node (SD pair), link‐sharing backup VPs are obtained among candidate VPs by information exchanges between the agents. In this scheme, the competition for a link between backup VPs is detected by the reception of messages passing the link, the function of collecting this detection information by an S node by information exchanges between agents is incorporated by each agent and the changes in the competition for the link can be followed. As an effect, it has a characteristic of high adaptability to the changes in the network resources (link capacity, topology) and resource requirements (primary VPs, backup VPs). In this paper, the results of performing simulation experiments to verify this characteristic are shown. The results of the experiments demonstrate that optimal backup VPs are obtained with good efficiency even when the number of backup VP candidates increases with the increase in SD pairs. In addition, the methods for obtaining the parameters of the agents required for obtaining optimal backup VPs are shown. © 2001 Scripta Technica, Electron Comm Jpn Pt 1, 85(3): 30–39, 2002