A Novel Algorithm for Fast Detection of Network Failure

Abstract

Fast failure detection is critical to traffic protection/restoration in a high-speed network. Generally, the way to detect failures is to periodically exchange keep-alive messages by layer 3 protocols. In traditional protocols like Open Shortest Path First (OSPF), the failure detection time is about tens of seconds, which makes the protection/restoration time too long to keep up with the increasing quality-of-service (QoS) demand. Although configuring the related parameters in OSPF can reduce the detection time to the order of milliseconds, it introduces notable issues of scalability and instability to the network. In this paper, a novel algorithm called “Low Priority Hello Reduction” (LPHR) is proposed to achieve faster failure detection while maintaining high network stability. This algorithm reduces the Hello messages sent when the signaling network is congested and maintains the adjacency by other messages. We compare it with previous schemes and our experimental results show that the algorithm we propose is superior in about an order of magnitude in reduction of both false failure alarms and queuing delay of other messages, especially at a light network load.