Lifetime-Aware Task Scheduling For Mixed-Critical Systems

Due to the development of integrated technologies,more and more embedded real-time systems integrate multiple functions on a single computing platform.This allows applications of different critical levels that originally run isolated to share processor resources.However,this kind of general-purpose computing platform not only needs to consider the cooperation between the functions,but also needs to consider the differences in critical levels of different applications.Therefore,the concept of mixed-critical systems is proposed.Mixed-critical systems are widely used in high-security areas such as automobiles and aerospace,but traditional optimization strategies for mixed-critical systems are designed to meet the critical levels of application programs,ignoring the optimization of system lifetime,resulting in the decay of system lifetime.However,frequent replacement of mixed-critical systems can result in expensive overhead.Therefore,to meet the critical level of application constraints,prolonging the service life of the mixed-critical system has a high research value.This paper aims to extend the lifetime of mixed-critical systems,consider both transient and permanent failures,and adopt a fault-tolerant solution to meet reliability requirements,so that the system can meet schedulability and reliability.This paper proposes a three-phase solution.Firstly,using dynamic voltage frequency scaling technique to determine the operating frequency for each task,to ensure that the thermal cycle has minimal damage to the equipment life.Secondly,using the re-execution technique,the system life is analyzed under the constraints of schedulability and reliability.And a multi-objective mixed-integer linear programming problem is obtained,then the re-execution numbers for each task is calculated.Finally,a lifetime-aware scheduling algorithm is designed according to the task's hot and cold characteristics.Simulation experiments show that compared with the other four benchmark experiments,the algorithm proposed in this paper is 27.3%,17.2%,18.2%,and 23.1% on average in terms of extended lifetime,on the premise of guaranteeing system scheduling and reliability.The increase is up to 38.1%,56.8%,52.5% and 43.6%.