机构地区: 华南理工大学
出 处: 《陕西科技大学学报(自然科学版)》 2007年第6期20-25,共6页
摘 要: 为了在Linux系统中实现对多媒体任务灵活有效的调度支持,作者引进并实现了多媒体任务模型ARB.与周期任务模型、零星任务模型等传统实时任务模型相比,对ARB任务模型做了如下几方面的改进:(1)不需要预先知道任务的计算资源需求量或最坏情况执行时间,以满足多媒体任务对资源需求动态多变的特点;(2)任务相邻两次执行之间需要满足一定的时间距离约束,以确保多媒体任务的执行延迟和抖动,使执行结果更为流畅;(3)多媒体任务以指定的速率执行,且该速率在任务执行过程中可以动态调整,以适应用户需求和系统可用资源的动态变化,在具体的内核调度算法实现中支持执行速率、截止期限、周期等重要实时任务属性的定义.实验结果表明,基于ARB,Linux内核可以为多媒体任务提供如下调度支持:(1)保证任务以指定的速率执行,且速率在任务执行过程中动态可调;(2)通过任务执行时的时间距离约束,保证任务的执行延迟和抖动;(3)通过准入控制和自适应QoS协商机制,实现资源的动态请求.因此,基于ARB任务模型,Linux内核不仅可以实现对多媒体任务灵活有效的调度支持,而且使得这些任务具有较好的可控性.这一点对实时任务很重要,有助于提高实时任务的可确定性,此外对于系统资源的合理分配以及充分利用具有重要意义. To support flexible and effective multimedia task scheduling in Linux kernel, an improved task model is introduced and implemented. Compared with the traditional real-time task models, such as periodic and aperiodic task models, the task model makes the following improvements: (1) there is no need for multimedia tasks to estimate or provide their execution time in advance; (2) time constraint between any two executions of one task must be satisfied to guarantee the required performance, such as delay and jitter; (3) multimedia tasks are guaranteed to make progress at the specified rate, which is run-time adjustable. Currently, the task model has been implemented in Linux kernel, and some important timing properties required by real-time tasks, such as deadline, period and execution rate, are supported. Experiments show that based on ARB, Linux kernel is capable of supporting multimedia tasks in this way: (1) they are guaranteed to make progress at the specified execution rate, which is run-time adjustable; (2) the run-time delay and jitter are guaranteed by the specified time constraint between any two adjacent executions of one task; (3) flexible and dynamic resource requests are possible by the admission control and adaptive QoS control mechanism. Therefore, based on the proposed task model, Linux kernel can provide more flexible and effective scheduling for multimedia tasks, and can make these tasks more controllable, which is significant not only to improve the determination of real-time tasks, but to make full use of system resource as well.
领 域: [自动化与计算机技术] [自动化与计算机技术]