Abstract: In order to effectively detect the failure of network system, a novel detection algorithm FDD-GS (Failure Detection algorithm based on Glowworm Swarm) is proposed by glowworm swarm. In this algorithm, the failure detection method and the minimum deviation optimization model have proposed by feature information entropy at first, and the algorithm processes is presented. Finally, an experiment with simulation platform was conducted with the study the key factors of FDD-GS. Compared to performance of other algorithm, the results show that FDD-FNN has low detection error.