Article Preview
Top1. Introduction
Real-Time systems are playing a crucial role in our society, and in the last two decades, there has been an explosive growth in the number of real-time systems being used in our daily lives and in industry production. Systems such as chemical and nuclear plant control, space missions, flight control systems, military systems, telecommunications, multimedia systems, and so on all make use of real-time technologies. The most important attribute of real-time systems is that the correctness of such systems depends on not only the computed results but also on the time at which results are produced. In other words, real-time systems have timing requirements that must be guaranteed. Scheduling and schedulability analysis enables these guarantees to be provided. Starting as an aid for industrially specialized embedded applications, Real time operating systems (RTOSs) are now common in a large variety of commercial products. Application areas are for example telecommunications, automotive, defense industry, medical equipment and consumer electronics. Common denominators for these embedded systems are real-time constraints. These systems are often safety critical and must react to the environment instantly on an event. Imagine for example the airbag of a car not going off instantly as a crash occurs; reaction time delay would be disastrous. Several interesting academic and industrial research works have been made last years to develop reconfigurable systems (Gehin & Staroswiecki, 2008).We distinguish in these works two reconfiguration policies: static and dynamic reconfigurations where static reconfigurations are applied off-line to apply changes before the system cold start (Angelov et al. 2005) whereas dynamic reconfigurations are applied dynamically at run-time. Two cases exist in the last policy: manual reconfigurations applied by user (Rooker et al. 2007) and automatic reconfigurations applied by Intelligent Agents (Khalgui, 2010); (Al-Safi & Vyatkin, 2007). In this paper, we are interested in the automatic reconfiguration of embedded real time Systems. We define at first time a new semantic of this type of reconfiguration where a crucial criterion to consider is the automatic improvement of the system’s feasibility at run-time. We propose thereafter an Agent-based architecture to handle all possible reconfiguration scenarios. Therefore, nowadays in industry, new generations of embedded real time systems are addressing new criteria as flexibility and agility. To reduce their cost, these systems should be changed and adapted to their environment without disturbances. It might therefore be interesting to study the temporal robustness of real-time system in the case of a reconfiguration where the reconfiguration results in a change of the value of tasks parameters: WCET, deadline and period. This new reconfiguration semantic is considered in our work and we will present its benefits. We are interested in this work in automatic reconfigurations of real-time embedded systems that should meet deadlines defined in user requirements (Baruah & Goossens, 2004). These systems are implemented sets of tasks that we assume independent, periodic and synchronous (e.g. they are simultaneously activated at time t = 0 time units). We assume also that the deadline of each task is equal to the corresponding period. Therefore the system’s implementation is dynamically changed and should meet all considered deadlines of the current combination of tasks. The schedulability of a task set with a given algorithm is said sustainable w.r.t. a parameter when a schedulable task set remains schedulable when this parameter is changed in a positive way. The sustainability is an important property since it permits to study the worst case scenario. In (Lakshmanan & Rajkumar, 2010), the authors prove that the critical scheduling instant characterization is easier in the context of sporadic real-time tasks. They provide, for systems scheduled under a rate-monotonic priority assignment rule, a pseudo-polynomial response-time test. The rest of the literature on self-suspending tasks focus on the multiprocessor context (Liu & Anderson, 2010). Our approach addresses the problem for periodic tasks, and is restricted to EDF and RM algorithms. We define an agent-based architecture that checks the system’s evolution and defines useful solutions when deadlines are not satisfied after each reconfiguration scenario and the Intelligent Agent handles the system resources in such way that, meeting deadlines is guaranteed. A tool RT - Reconfiguration is developed and tested in our laboratory to support the agent’s services.