Function Allocation and Bandwidth Reservation for Mixed-critical Adaptive Software Systems

Abstract

The new AutoSAR adaptive platform makes mixed-critical automotive systems able to adapt themselves in response to hardware and software failures at runtime. However, mapping functions of these automotive systems and reserving bandwidth for them are still major challenges. In this paper, we propose a model-based approach for mapping functions of an automotive system to its hardware nodes and reserving their bandwidth. To do so, an architecture description language for automotive systems (i.e. EAST-ADL) is used to design an embedded system, and to specify its timing requirements. The design model is then used for identifying functions allocation and their bandwidth in different system configurations. To schedule the critical functions of the system, the Earliest Deadline First (EDF) is used, while the Constant Bandwidth Server (CBS) is used for scheduling the non-critical functions. The quality of service for the non-critical functions is determined by their reserved bandwidth. In addition, a Tabu search-based approach is used for mapping the system functions to hardware nodes. Furthermore, there is a temporal isolation between the critical and non-critical functions. Thus, overruns of the non-critical functions do not affect the timing guarantees of the critical functions, and the quality of service for the non-critical functions is maximized.