A Simulation Framework for Fault-Tolerant Clock Synchronization in Industrial Automation Networks Fritz Praus, Wolfgang Granzer, Georg Gaderer, Thilo Sauter Research Unit for Integrated Sensors Systems, Austrian Academy of Sciences Viktor Kaplan Strasse 2, A-2700 Wiener Neustadt, Austria {Fritz.Praus,Wolfgang.Granzer,Georg.Gaderer,Thilo.Sauter}@oeaw.ac.at Abstract Many applications such as distributed measurements or real-time networks benefit from a common notion of time. Protocols providing high precision and simple clock synchronization are necessary to achieve such a com- mon time base. However, most of the available proto- cols are lacking with regard to fault tolerance and per- formance in case of a fault. The project IMAGINE (Intro- duction of Master Group Based Industrial Ethernet) over- comes these limitations by introducing a fault-tolerant IEEE1588 master group. A proof of concept for a large- or even medium- scale network is, however, very difficult to obtain under laboratory conditions. Therefore, a simu- lation framework has been developed, which is presented in this paper. 1. Introduction Synchronization of processes is an essential require- ment in distributed systems. One possibility to do so is clock synchronization, and various technologies to syn- chronize clocks exist today. Key properties of a clock synchronization approach are its achievable precision and accuracy. Precision refers to the synchronization of the clocks with respect to each other whereas accuracy is de- fined as synchronization of nodes with respect to an ex- ternal reference (e.g., GPS time). Currently, an accuracy of nanoseconds or even better is possible. Still, it is desir- able that these highly precise clock synchronization pro- tocols are kept as simple and robust as possible. On the one hand, the protocol shall be fault-tolerant in that syn- chronization must be possible in a reasonable time even when key components of the network (e.g, a synchroniza- tion master) fail. On the other hand, resources require- ments (e.g., regarding network bandwidth, node capacity) shall be as small as possible especially in large networks. As shown in Section 2, none of the available clock synchronization protocols satisfy the above mentioned re- quirements. The research project IMAGINE (Introduc- tion of Master Group Based Industrial Ethernet) 1 tries to overcome these problems by introducing a new approach. Particular attention is paid to improving the robustness of standard synchronization methods popular in automation. Evaluating and judging these newly defined algorithms raises the question of performance analysis under realistic operating conditions. Analytical methods allow to describe problems in a very exact, provable, and predictable form. However, it is very difficult or nearly impossible to analyze the dy- namical behavior of complex systems. Quite often, it is only feasible to describe steady states of such systems. Prototyping allows to study the real-world behavior of various systems. Results of such an analysis can be achieved very fast and the basic functionality of a system can be shown. It may however be difficult to perform such an analysis under laboratory conditions in a reproducible way. Moreover, large-scale analysis with more than a few participating nodes is neither easily manageable, nor can it be implemented in a cost-efficient way. Simulation is an approach mainly targeted to analyz- ing dynamical systems. Abstract models of a particular system are developed and evaluated using a simulator. It is thus possible to simulate large-scale installations con- sisting of thousands of nodes. Compared to prototyping, a simulation is not bound to real-time. To simulate the behavior of a system over years, a simulator needs only a fraction of time that would be needed using prototyp- ing. Obviously, the effective duration of the simulation depends on the available system resources (i.e., compu- tational power and memory size) of the simulation host system as well as on the level of detail of the simulation model. For the performance analysis of large-scale and hetero- geneous systems which are within the focus of IMAGINE, simulation is the method of choice. Nevertheless, the se- lection and implementation of suitable simulation strate- gies is by no means straightforward. The purpose of the paper is thus to discuss the simulation framework which has been developed within the scope of the project. 1 The work presented in this paper was funded by the Austrian Re- search Promotion Agency (FFG) under the BRIDGE project 810092. 1-4244-0826-1/07/$20.00 © 2007 IEEE 1465