QoS ANALYSIS AND COMPARISON OF FAILURE DETECTION ALGORITHMS

Abstract

With the increasing trend of distributed computing, there is greater need of increasing its reliability and ' dependability by providing fault tolerance. Failure detection is an important aspect of fault tolerant systems. To make the failure detector itself reliable, it should satisfy some Quality of Service metrics which are quantified by speed and accuracy of the failure detector. The earlier works try to quantify the failure detection process with QoS metrics which are implementation specific and so they cannot be not be used for comparing different failure detection algorithms. In this dissertation work, failure detection in multicomputer systems by using heartbeat messages is studied and implementation independent QoS metrics for quantifying the failure detection process are given. Two failure detection algorithms that use heartbeat mechanism for failure detection are compared with respect to the proposed Quality of Service (QoS) metrics. A configuration procedure to make the failure detector adaptive to the network traffic condition is discussed which sets the parameters of the failure detector. A generalized simulation framework has been developed in C++ and the results for comparison of two algorithms and configurator outputs are presented and discussed.