Performance Analysis of Hierarchical Failure Detector

We present a new failure detector implementation. This implementation, a variant of the heartbeat failure detector, is both adaptable and designed for scalability. Its first specificity of our implementation lies in the fact that it is designed as a shared service among several applications by way of an adaptation layer between the failure detector and the application. This layer adapts the quality of service according to application needs. The second specificity is the hierarchic organization of the detection service: it allows to decrease the number of message and the processor load. Through an experimentation evaluation, we show that our implementation is adaptable to the environment characteristics and usable with large scale applications.