A simulation model for event goodput estimation in real-time sensor networks

Abstract

In this paper we propose event goodput, i.e., the fraction of events which may be successfully managed by a system, as a relevant metric to describe the performance of battery powered real-time sensor networks. Unlike other performance metrics as response, completion, maximum lateness times, all representing fundamental, but different, figures of merit for the description of the behavior of real-time systems, event goodput provides an immediate and a direct relation with the events which may be satisfactorily managed (or not) by a real-time application. We will show such metric well serves the purpose of describing the performance of battery powered, random event-driven networks, such as sensor networks deployed for surveillance and intrusion detection applications, operating in time critical scenarios. In essence, such real-time systems may be assessed in terms of the fraction of events which they successfully/unsuccessfully detect and report within a time interval of interest. The importance of such metric is here demonstrated providing a simulation model and results where the use of the event goodput metric is discussed in conjunction with those metrics which are traditionally utilized for the assessment of a real-time sensor networks.