High precision testbed to evaluate ethernet performance for in-car networks

Abstract

Validating safety-critical real-time systems such as in-car networks often involves a model-based performance analysis of the network. An important issue performing such analysis is to provide precise model parameters, matching the actual equipment. One way to obtain such parameters is to derive them by measurements of the equipment. In this work we describe the design of a testbed enabling active measurements on up to 1 [Gb/Sec] Copper based Ethernet Switches. By use of the testbed it self, we conduct a series of tests where the precision of the testbed is estimated. We find a maximum error of ±55 [ns] measuring frame traversal time, and quantify the constant error imposed by the testbed it self.