Experimental evaluation on optimality of multiagent-based backup VP reservation method

In order to enhance the reliability of ATM networks, recently (Y. Kakuda, 1998), the virtual path (VP) reservation method using multiple agents (MA) has been proposed. In this MA method, given a primary VP and multiple backup VP candidates between each source node/destination node pair, then for each such pair, a backup VP is selected from the given multiple backup VP candidates, and a primary VP and a backup VP are reserved so that, when some nodes and links in the primary VPs fail, the primary VPs are able to be changed to the corresponding backup VPs rapidly. This paper describes an experimental evaluation of the MA method with respect to optimality. The experimental results show that the MA method has good optimality, along with an increase in the number of backup VP candidates.