In a multi-ring network with a plurality of rings to which RPR is applied connected, high-speed failure recovery is enabled when a failure occurs between the rings. The contact nodes 700-1 and 700-3 connected to the living inter-link 505 periodically transmit and receive a KeepAlive signal to/from each other. Then, when a KeepAlive signal fails to arrive for a fixed time period, determination is made that the inter-link 505 develops a fault to broadcast-transmit an inter-link failure notification RPR-MAC frame indicative of failure occurrence on the inter-link to each node. A contact node as a transmission source of the inter-link failure notification RPR-MAC frame and each node having received the inter-link failure notification RPR-MAC frame immediately flush learning contents of their own learning data bases. A contact node connected to the spare inter-link 506 switches the spare inter-link 506 to a forwarding state.