A novel self-regulatory failure detection algorithm for distributed storage system

Distributed storage systems have the advantages of self-organizing, scalability, robust and fault-tolerant, which provide fundamental support for mission-critical applications. Unfortunately, there still exist some problems in real network conditions, especially, when considering asynchronous distributed environments, where messages may be delayed indefinitely and nodes may fail. So, failure detector is one of basic components to build a reliable distributed storage system. Considering the highly dynamic characteristics of distributed system, a novel self-regulatory failure detection algorithm (SFDA), which can combine heartbeat strategy with unbiased grey prediction model, is designed to improve the failure detection quality of system (QoS) according to the application needs and network environment changes. The key goal of using correction value in SFDA is to provide adoptable and usable service in all kinds of applications. The results show that, failure detector based on SFDA is more effective and flexible than other failure detectors.