Research On Mixed-criticality Task Scheduling Strategy Based On Multi-core System

In recent years,mixed-criticality systems have become a hot research topic in embedded real-time systems.In order to meet the needs of gradual diversification and complexity,how to improve the service quality of low-criticality tasks has been the focus of academic circles.At the same time,mixed-criticality systems were deployed to multi-core systems with higher performance,but multi-core systems required greater energy consumption.How to reduce the energy consumption while ensuring the original reliability of the system also became difficult.Therefore,studying the scheduling strategy and energy saving strategy of mixed-criticality systems has important practical effects and significance.This paper focuses on single-core platforms and multi-core platforms,and conducts related research on mixed-criticality systems,which are mainly divided into two parts:(1)research on mixed-criticality task scheduling strategies;(2)research on energy saving strategies for mixed-criticality systems.This paper first proposes a mixed-criticality task scheduling strategy DDTS based on dynamic degradation for a single-core platform.The difference from the fair degradation policy is that the strategy dynamically determines the service level of lowcriticality tasks based on the number of high-criticality tasks overrun.The higher number of high-criticality tasks that exceed the limit,the lower service level of lowcriticality tasks,so as to improve the service quality of low-criticality tasks.Experiments show that the DDTS strategy improves the service quality of lowcriticality tasks than the comparison strategy 10% or more.Then this paper extends the DDTS strategy to multi-core platforms,and proposes a multi-core mixed-criticality task scheduling strategy LBTS based on load balancing.This strategy uses a round-robin task division strategy in the task division phase,and DDTS strategy in the single-core scheduling phase.To achieve the purpose of improving the service quality and load balancing of low-criticality tasks,experiments show that this strategy not only makes the multi-core processor load more balanced,but also improves the service quality of low-criticality tasks by about 16.5%.Subsequently,this paper proposes a dynamic stretch-based mixed-criticality energy-saving strategy DSES for a single-core platform.This strategy uses DVFS technology on low-criticality tasks to stretch its execution time and deadline while maintaining its utilization rate.According to the different release cycles after the lowcriticality tasks are downgraded,the available slack time is allocated to reduce the energy consumption of the single-core platform system.Experiments show that this strategy saves about 15.3% in the single-core system compared with the comparative strategy.Then this paper proposes a multi-core mixed-criticality energy-saving strategy based on slack reclamation and task replication for multi-core platforms.This strategy uses task replication technology to schedule multiple copies of the same high-criticality task on multiple processor cores,based on the actual execution of high-criticality tasks,appropriately canceling its master copy tasks,dynamically recovering the slack time in the system,and allocating slack time based on the actual utilization of the service level of the low-criticality tasks,As a result,the execution rate of low-criticality tasks can be changed to maintain the original reliability of the system and reduce the energy consumption of the system.Experiments show that the SRTR strategy saves about 17.2%energy compared to the comparative strategy.