Efficient Network Fault Detection Using Adaptive Polling

Abstract

Network Management Stations (NMSs) poll the discovered entities in a network topology to determine their operational status classifying them as faulty or normal. This fault detection strategy is prone to scalability issues resulting in delayed fault assessment and reporting, especially when monitoring large network topologies. We study network fault characteristics in light of network baseline statistics and propose a novel adaptive polling strategy. The proposed polling strategy takes into account past behavior of the network and utilizes this information to dynamically formulate polling strategies to achieve faster detection of network faults, while regulating the amount of network management traffic generated. Simulation results show that this leads to lower Mean-Time-to-Detect (MTTD) for network faults compared to traditional network management solutions, which employ static pre-configured polling strategies.