【作者】 李龙；

【导师】 李仁发；

【作者基本信息】 湖南大学 ， 计算机科学与技术， 2014， 硕士

【摘要】 在汽车电子系统、航空电子系统等安全性关键（Safety-Critical）实时嵌入式系统中,随着应用数量不断增加,系统的结构也越来越复杂,为了提升性能、节约成本、功能多样化以及安全可靠等方面的考虑,系统将越来越多的不同关键级的功能单元集成到了一些共享硬件上。在这样的混合关键级系统中,不同关键级的任务在同一共享硬件资源下进行调度会存在任务间相互干扰和影响,给任务调度带来很多新的问题。当前混合关键级系统调度算法研究存在调度模型不完善和系统设计不灵活等方面的不足。本文针对这些问题,通过使用理论与实践相结合的方法,对混合关键级系统实时调度算法进行研究。本文主要研究工作如下:首先,针对现有混合关键级任务调度算法一般只考虑双关键级,无法很好地应用到关键级数大于两个关键级的实时系统中,本文对现有响应时间分析进行了改进,将其从双关键级扩展到多关键级,提出了基于响应时间分析的支持多个关键级的实时任务调度算法。其次,鉴于当前混合关键级任务调度算法在系统发生模式切换时为了保证高关键级任务执行会立即抛弃相对低关键级任务,对于低关键级任务的处理过于消极,本文提出了基于弹性调度模型的调度算法,积极的处理低关键级任务。实验表明该算法在保证高关键级任务正确执行同时使得混合关键级系统可接受任务数目提升了32%左右。最后,现有混合关键级调度算法研究中,大都是基于仿真平台做实验,并没有考虑到系统的真实开销（如任务迁移开销、系统调度开销）,而现有实时操作系统内核并没有支持关键级的调度算法,因此本文对任务调度中的系统开销进行详细分析总结,在此基础上提出了蕴含开销的调度测试算法,并基于LITMUS^RT框架在Linux操作系统上搭建起混合关键级调度算法实验平台,将理论与实践很好的结合在一起。更多还原

【Abstract】 In modern safety-critical real-time embedded systems(such as automotive electronic systems,avionics systems),along with the increasing number of applications,the structure of the systems are more and more complex,considering performance,cost savings,functional diversification,safety and reliability,an increasingly important trend of these systems is the integration of componets with different levels of criticality onto a common hardware platform.In such a "mixed-criticality" system,different task with different criticality may interference and influence each other which would bring a lot of new scheduling problems.There are a lot of deficiencies exist in the current research on mixedcriticality system scheduling area(such as inadequate scheduling model,inflexible system design).In order to solve these problems,we use a method which is the combination of theory and practice to study the mixed-criticality real-time scheduling algorithm.The main research works are as follows:Firstly,taking into account that the most existing mixed-criticality scheduling algorithms only consider the case of dual-criticality systems,we introduce a new response time analysis algorithm which support for multi-criticality systems.Secondly,considering the current mixed-criticality scheduling algorithm will abandon the relatively low-critical tasks immediately in order to ensure the higher critical task execution when the system criticality increases,it is too negative,this paper based on the elastic scheduling model,proposed a elastic scheduling algorithm for mixed-criticality systems,handling the low-critical tasks positively.Experiments show that the algorithm is significantly improving the number of acceptable tasks.Finally,the existing scheduling algorithms studies mostly based on simulation experiment platform,such experiment results are too idealistic and did not considering the real system overhead(such as task migration overhead,system scheduling overhead).Therefore,this paper based on detailed analysis of the scheduling overhead,proposed an overhead-aware scheduling model,and implement the experiment platform on the Linux operating system,the scheduling algorithm is improved through the feedback of scheduling algorithm in the actual running,the overhead-aware experiment platform erected the bridges between theory and practice,which greatly enhance the practicability of the scheduling algorithm.更多还原