Research Of A QoS-based Adaptive Failure Detection Algorithm

In distribution system, failure detector is an important protective technology and the base of the functions for the high usability.It also has important applications in the fields of grid computing, cluster management, communication protocol, network of things, cloud computing, etc.This dissertation has deeply studied the QoS of failure detection model and algorithm improvement. The main achievements of this dissertation can be described as follows:1. In according to Quality-of-service (QoS) of failure detectors, a relatively complete set of QoS evaluation system is proposed in this paper on the basis of previous studies. Chen proposed a set of metrics to evaluate the QoS of failure detectors. However, Chen only discussed the QoS metrics of failure detectors in terms of the failure detection time and the accuracy properties. In addition, Chen et al. showed that, the QoS of failure detectors can be captured by TD, TMR, TM, which can be the primary QoS metrics. All of the other QoS metrics can be derived from these three QoS metrics. After analyzing the previously published QoS metrics, a new QoS model is proposed in this paper, which consists of completeness, accuracy and timeliness.2. A new failure detection algorithm based on QoS is proposed in this paper after researching and analyzing the existing adaptive failure detection algorithms. Power low is used to calculate mean value, predicting message delay in exponential distribution. Adaptive failure detection requires that failure detectors can dynamically adjust the detecting quality according to the variations of applications and the load of the network. Because of the instability of the network, old failure detectors without adaptive mechanism can not meet the requirements of QoS (Quality of Service) of applications all the time. Detect speed and accuracy is the crux of failure detector. A QoS-AFD adaptive failure detection algorithm is realized. Experimental results and analysis show that detect time and the accuracy of QoS-AFD are relatively high.