Dependability Metrics for Network Systems—Analytical and Experimental Analysis

Paper presents several dependability metrics for network systems. Methods of metric calculation are also taken into consideration. Two general groups of metrics are analyzed: analytical and experimental. Following analytical metrics are analyzed: network cohesion in the system, number of services involved in the compound service, absolute importance, dependence and criticality of the service and overall reliability of the compound service. Experimental metrics—task response time and service component availability—are also calculated. Numerical experiments were performed on a test case scenario. The dependability means the combination of the reliability and functional parameters of the network. Analysis is done with a usage of open-source simulation environment that can be easily modified and extended for further work. Based on the simulation results, some alternatives can be chosen in case of system or service failure. This way it is possible to operate with large and complex networks described by various—not only classic—distributions and set of parameters. The presented problem is practically essential for organization of network systems.