OneOS的任务调度系统设计与实现

Design and Implementation of OneOS Task Scheduling System

在物联网应用场景下,操作系统应该保证对关键事件作出实时响应,同时,还应该提供丰富的任务管理功能,以便应对各种上层应用需求,为此提出了关键事件处理的性能分析模型,并根据此模型分析出任务调度系统性能改善的主要方向。OneOS使用混合调度算法、优先级位图算法、前置计算待调度任务算法、哈希桶算法改善调度性能;同时,通过加锁粒度的优化,有效缩短了最大中断响应时间,进一步保证了处理关键事件的任务(最高优先级)能够快速响应。最后,实验证实了OneOS任务调度系统在同类产品中对关键事件的响应性能得到了优化。

In the application scenario of the internet of things,the operating system should ensure real-time response to key events.At the same time,it should also provide rich task management functions to meet various upper application requirements.In the paper,a performance analysis model of key event processing is proposed,and according to this model,the main direction of improving the performance of task scheduling system is analyzed.OneOS uses hybrid scheduling algorithm,priority bitmap algorithm,pre-computing task to be scheduled algorithm and hash bucket algorithm to improve scheduling performance.At the same time,through the optimization of lock granularity,the maximum interrupt response time is effectively reduced,and the task(the highest priority)that handles key events can respond quickly.Finally,the performance comparison data are given,which proves that the response performance of OneOS task scheduling system to key events has been optimized in similar products.