For reducing routing information to achieve scalability in large ATM networks, ATM Private Network-to-Network Interface (PNNI) adopts hierarchical routing. In order to provide efficient routing, a large ATM network is decomposed into subnetworks, called peer groups (PG), and peer groups will advertise aggregated routing information only. Because the aggregated information is not very precise, a call set up message may be rejected on a chosen route. When an ATM node discovers a call set up message cannot proceed due to insufficient resource, it initiates a back-tracking procedure called "crankback" against call blocking. Although crankback reduces blocking probability, it also causes additional overhead, such as longer setup delay. Therefore, in this paper, we propose two approaches to reduce crankback overhead. The first approach adds additional information into the setup message, referred to as CIS (Crankback Information Stack), to reduce crankback overhead. The other approach, referred to as CT (Cost Threshold), uses aggregated path cost and the cost information of previously rejected call setup message to determine whether call setup on the next alternate path should be tried. Our simulation results show that both of the proposed approaches reduce crankback overhead significantly. Especially, the combination of CIS and CT approach achieves further improvement.
