Research Of Hierarchical Real-time Scheduling Algorithm Based On TDM

In the field of real time embedding system,hierarchical scheduling frameworks have received more and more attention because of the variety and complex of the task models in real time scheduling.Due to the simpleness of implement and stability,the TDM-based hierarchical scheduling policy plays a dominant role in such frameworks.In resent years,the notion of mixed-criticality task model caused a great sensation and it has been one of the main research in the field of real-time scheduling immediately.Now-days,a great number of researches has been done on the basis of mixed-criticality task model.Meanwhile,these researches are all most using an integrated manner to schedule mixed-criticality systems and get a better utilization of system resource(a high acceptance ratio in experiments).However,the intergraded scheduling approach requires a centric scheduler to manage all the tasks crossing different criticality levels,which conflicts with the common requirement of strict functional isolation in mixed-criticality system design.And in the design of mixed-criticality systems,the only system architecture system designers can accept is based on virtualization,where the subsystems with different criticality levels are wrapped into different virtual machines and all the virtual machines are managed by a hypervisor.That way,the logical interaction between subsystems with different criticality levels is minimized.As a result,using a hierarchical scheduling approach to deal with the scheduling and analysis problem in mixed-criticality systems is a logical trending.This thesis used a hierarchical scheduling framework based on TDM strategy to schedule mixed-criticality real-time systems in the presence of functional isolation requirement which prevent the interference between subsystems crossing different criticality levels.Further,this thesis presented two methods of analyzing the schedulability of hierarchical scheduling framework based on TDM strategy and introduced how to determine the resource allocation to different task sets crossing criticality levels to optimize the system schedulability.Experiments with synthetic workloads show that the scheduling approach presented in this thesis can greatly improve the system schedulability comparing with the isolated approach.