Collaborative Research: CNS Core: Medium: Parallel and Real-Time Multicore Scheduling for an Efficiently-Used Cache (PARSEC)

合作研究：CNS 核心：中：高效使用缓存的并行实时多核调度 (PARSEC)

• 批准号: 2211641
• 负责人: Nathan Fisher
• 金额: $ 29.51万
• 依托单位: Wayne State University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-01 至 2026-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2211641&HistoricalAwards=false
• 关键词: Collaborative; Research; CNS; Core; Medium

Safety-critical systems that have strict “real-time” requirements are becoming increasingly ubiquitous and complex. Epitomizing this recent trend toward sophisticated real-time systems are autonomous vehicles, which must perform image recognition, machine learning, routing, and planning tasks, simultaneously and with minimal delay. Furthermore, these real-time computational tasks must execute upon shared hardware (e.g., processors, memory, storage) due to the severe constraints on the size, weight, and power of the entire system; however, the sharing of computer resources creates tremendous contention and competition between tasks. This project addresses a fundamental challenge of how multiple real-time, safety-critical tasks can effectively share the underlying memory architecture and still meet timing constraints. In particular, this project will develop a novel system design and analysis framework called PARSEC (Parallel and Real-Time Multicore Scheduling for an Efficiently-Used Cache). PARSEC contributes to the state-of-the-art with (a) new multicore scheduling algorithms that explicitly manage how contending tasks share memory resources; (b) new formal analysis techniques that verify that a system’s timing constraints are satisfied with existing memory resources; and (c) a set of open-source automated tools that will enable system designers to utilize the framework on commercial off-the-shelf processing architectures. PARSEC will be implemented and evaluated upon the popular RISC V architecture to facilitate wide dissemination to the public.This project will result in safer, more efficient designs of time-sensitive systems, including autonomous vehicles and robotics. Furthermore, the resulting research and system design techniques in this project can be applied to any real-time, safety-critical systems executing concurrent computational tasks upon a shared processor and memory. The reduction in contention in the memory hierarchy obtained from project artifacts will potentially lessen demands on power and fuel in safety-critical systems, decreasing their carbon footprint. The project will benefit the educational missions of University of Nevada Las Vegas and Wayne State University by providing a unique training, education, and experiential learning opportunity for undergraduate and graduate students via course projects related to safety-critical system design. To aid other researchers, this project will also disseminate research results through publications, public talks, tutorials, project websites, and online videos.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

具有严格“实时”要求的安全关键系统正变得越来越普遍和复杂。自动驾驶汽车是复杂实时系统这一最新趋势的缩影，它必须以最小的延迟同时执行图像识别、机器学习、路由和规划任务。此外，由于整个系统的尺寸、重量和功率受到严格限制，这些实时计算任务必须在共享硬件（例如处理器、内存、存储）上执行；然而，计算机资源的共享在任务之间造成了巨大的争用和竞争。该项目解决了多个实时、安全关键任务如何有效共享底层内存架构并仍然满足时序限制的基本挑战。特别是，该项目将开发一种新颖的系统设计和分析框架，称为 PARSEC（高效使用缓存的并行实时多核调度）。 PARSEC 通过以下方式为最先进的技术做出贡献：(a) 新的多核调度算法，该算法显式管理竞争任务如何共享内存资源； (b) 新的形式分析技术，用于验证现有内存资源是否满足系统的时序约束； (c) 一套开源自动化工具，使系统设计人员能够利用商业现成处理架构的框架。 PARSEC 将在流行的 RISC V 架构上实施和评估，以促进向公众的广泛传播。该项目将带来更安全、更高效的时间敏感系统设计，包括自动驾驶车辆和机器人。此外，该项目中产生的研究和系统设计技术可以应用于在共享处理器和内存上执行并发计算任务的任何实时、安全关键系统。从项目工件中获得的内存层次结构争用的减少将有可能减少安全关键系统中对电力和燃料的需求，从而减少其碳足迹。该项目将通过与安全关键系统设计相关的课程项目，为本科生和研究生提供独特的培训、教育和体验式学习机会，从而有利于内华达大学拉斯维加斯分校和韦恩州立大学的教育使命。为了帮助其他研究人员，该项目还将通过出版物、公开演讲、教程、项目网站和在线视频传播研究成果。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。