支持SRP协议的实时调度技术

Real-Time Scheduling Technique Under SRP Protocol

实时操作系统对多任务资源访问控制提出了较高的要求.资源竞争常会引起优先级翻转问题导致任务阻塞,增加资源等待延迟时间.传统资源访问控制协议可以部分解决此问题,但是存在上下文切换次数高,任务调度效率低等不足.在结合SRP协议的提前阻塞特性和Mini Core实时调度器设计后,分析了协议中任务调度规则,指出了调度器支持SRP协议时效率低下的原因,并引入胜者树结构改进就绪队列,给出了新的任务搜索算法.理论分析与实验结果表明改进的调度器在任务集规模较大时,提高了调度效率,较好的支持了SRP协议.

Real-time operating systems lay claim to multitasking resource accessing control. Resources competition not only results in priority inversion and tasks blocking, but also prolongs timing delay caused by acquiring resources regularly. Although conventional resources accessing protocol works out priority inversion issues, there are still some defects such as frequent context switch, scheduling efficiency etc. With a research on the implement of Mini Core OS and SRP, this paper analyses the detailed criteria of scheduling and demonstrates the causation of low scheduling efficiency under SRP. To solve these issues above, we proposed and implemented a ready queue based on winner tree and a job selection mechanism with low timing overhead. Theoretical analyses and experiment results justify the better performance on large scale tasks scheduling under SRP protocol.