Scheduling and Fault-Tolerance with Free Open-Source Components for Real-Time Applications M. Silly-Chetto 1 , A. Marchand 1 , T. Garcia 1 and C. Plot 3 1 IRIN / University of Nantes Rue Christian Pauc 44306 Nantes cedex 03 France maryline.silly@iut-nantes.univ-nantes.fr 2 CRTTI / IUT of Nantes La Chantrerie CP3003 44087 Nantes cedex 03 France plot@iut-nantes.univ-nantes.fr In this paper, we will describe a national project (work supported by the French research office) that aims at the improvement of embedded compu- ting systems for applications with real-time constraints. The objective of this project is first to create a library of free software components for the design of real-time operating systems and second, to participate in the evolution of an opened community standard, Linux. The key objective is as well to demons- trate the applicability and the interoperability of these software components by simulation, by integration and finally by tests on a real application. The tests on a mobile robotic platform (an Automated Guided Vehicle) are per- formed to show the benefits in terms of both improved integration process and adequacy with strict requirements on safety and reliability of next generation applications. First, strategies to equip Linux with real-time are presented and illustrated with the open-source real-time system LINUX/RTAI. Then, we are interested in the CLEOPATRE project itself and more precisely in its ob- jectives in terms of providing dynamic scheduling and timing fault-tolerance. In addition, we tackle the problem of overload management with a brief des- cription of the existing scheduling schemes before introducing the library of free software components. Finally, we present the CLEOPATRE demonstrator used to prove the efficiency of the new Linux-based operating system. 1 Real-Time under Linux 1.1 The Open-source model The open-source sotftware concept was born, there is a score of years, in the academic research centers. Since its beginning, the research community has observed an exponential evolution of the number of users mostly thanks