Research And Implementation Of Mixed Criticality Tasks Scheduling Experimental Platform On Multi-core Systems

Multi-core processor architecture has become a primary means of improving processor performance and gradually applied to the design of real-time systems. An increasing trend in embedded system design is to integrate a number of relatively independent subsystems with different levels of criticality into a shared hardware platform for better cost, weight and power efficiency. Using traditional real-time scheduling techniques to schedule mixed-criticality systems may be unable to efficiently utilize the computational bandwidth provided by multi-core platforms and result in unacceptable resource waste. This is mainly due to the large gap between worst-case execution time prediction in schedulability analysis and actual execution time at runtime. So after mixed-criticality task model was proposed, in the area of real-time systems research on schedulability of mixed-criticality system quickly became a hot issue.Even on a single-processor platform scheduling analysis of such a mixed-criticality system is also very challenging, in a multi-processor platform will be more difficult. There is currently no operating system support mixed-criticality real-time task model, restricting its practical application in the field of real-time systems, therefore an urgent need to establish real-time operating system experimental platform for test different mixed-criticality real-time scheduling algorithm runtime performance.In this thesis, the hierarchical scheduler of supporting mixed-criticality real-time model on multi-core platforms, is designed and implemented based on Linux kernel 3.0.0 with integrated Litmus 2012-3 patch. Support the integration of different mixed-criticality scheduling algorithm. Meanwhile implement the criticality monotonic priority assignment for mixed-criticality hierarchical scheduling on multi-core platforms, task at each criticality level are scheduled by different intar-container schedulers. The four key technologies have been studied and developed, which include the real-time task runtime state transitions and dynamic preemption, the efficient runtime CPU priority queue management, efficient management of real-time task release queues and ready queues, and the ghost job status dynamic monitoring and processing during the entire life cycle of real-time routines. The real-time application management system has been developed to making the design and creation separated from the executing on mixd-criticality real-time task. The correctness and effectiveness of the hierarchical scheduler has been demonstrated by sufficient experiments. This work enables run-time performance analysis and comparison between more multi-core mixed-criticality scheduling algorithms, and will promote the application of the mixed-criticality scheduling strategy in practical real-time systems.