CPS: Breakthrough: Collaborative Research: Bringing the Multicore Revolution to Safety-Critical Cyber-Physical Systems

多核平台有可能使嵌入式网络物理系统的功能发生革命性变化。不幸的是，当这类系统具有安全关键型组件时，多核平台很少使用。原因是缺乏与核心之间共享的硬件组件(例如高速缓存、存储器控制器等)相关联的可预测性。在目前的技术中，必须对这些共享资源的使用做出非常保守的估计，以确保在运行时不会发生过度使用违规。由此产生的过度配置可能会非常严重，很容易否定任何额外核心的处理能力。该项目的目标是通过开发能够以可预测的方式控制共享硬件资源的分配机制来解决这一多核“可预测性问题”。该项目的研究议程包括相关实时资源分配问题的基础研究，涉及实时操作系统和中间件的原型工作，以及对所开发的定时分析工具(用于确定任务执行-时间预算)中所开发机制所带来的改进的实验评估。解决与多核平台相关的“可预测性问题”将是航空电子和汽车等领域安全关键的网络物理系统的突破性成果。当使用多核平台在这些域中托管高度关键的工作负载时，当前的技术状态是通过关闭除一个核之外的所有核来避免可预测性问题。除非找到更智能的解决方案，否则这些域不会从多核平台提供的大小、重量和功耗(交换)的节省以及功能的提高中受益。更广泛的影响包括与业界同行就支持无人驾驶飞行器的实时工作量进行联合研究，开发可供其他机构用于研究和教学目的的公开可用的开源软件，以及开发一门关于网络物理系统的新课程。

Multicore platforms have the potential of revolutionizing the capabilities of embedded cyber-physical systems. Unfortunately, when such systems have safety-critical components, multicore platforms are rarely used. The reason is a lack of predictability associated with hardware components such as caches, memory controllers, etc., that are shared among cores. With current technology, very conservative estimates concerning the usage of these shared resources must be made, to certify that overuse violations do not occur at runtime. The resulting over-provisioning can be significant, easily negating the processing power of any additional cores. The goal of this project is to resolve this multicore "predictability problem" by developing allocation mechanisms that enable shared hardware resources to be controlled in a predictable way. The research agenda in this project includes fundamental research on relevant real-time resource allocation problems, prototyping efforts involving real-time operating systems and middleware, and experimental evaluations of improvements enabled by the developed mechanisms in timing analysis tools (which are used to determine task execution-time budgets).Addressing the "predictability problem" associated with multicore platforms would be a breakthrough result for safety-critical, cyber-physical systems in domains such as avionics and automobiles. When using multicore platforms to host highly-critical workloads in these domains, the current state of the art is to obviate the predictability problem by turning off all but one core. Unless a more intelligent solution can be found, such domains will not benefit from savings in size, weight, and power (SWaP) and gains in functionality that multicore platforms afford. Broader impacts include joint research with industry colleagues on supporting real-time workloads in unmanned air vehicles, the development of publicly-available open-source software that can be used by other institutions for research and teaching purposes, and the development of a new course on cyber-physical systems.