A Network Application Fault Repair Method Based on Reinforcement Learning

Abstract

With the continuous expansion of enterprise networks and the growing emphasis on security requirements, many enterprises have implemented multi-level security equipment (MLSE) within their networks. However, this deployment has led to network application failures and reduced overall network availability. This paper conducts an analysis of the underlying mechanisms responsible for such network application failures in MLSE enterprise networks. Subsequently, a Fault Location and Recovery (FDR) method is proposed, utilizing Q-learning path search with an experience storage (ESQL) mechanism. To validate the proposed approach, a prototype MLSE enterprise network system is constructed using NFV technology, and a series of verification experiments are performed on the FDR method. The experimental results demonstrate that the proposed approach successfully reproduces the failure phenomena and accurately locates and expedites the repair of network application failures through the FDR method.