CSR: Small: Formal Foundations of Certifiable Mixed-criticality Systems

本研究解决了嵌入式系统中两个重要趋势的融合所产生的问题。首先，许多安全关键型应用程序都要遵守认证要求。第二，集成体系结构越来越趋向于在单个计算平台上支持多种功能（通常是不同的关键功能）。随着这些系统变得越来越复杂，获得所需的认证变得越来越具有挑战性。本项目研究以下论题：当前形式的调度理论不适合设计这样的混合临界（MC）系统，这些系统受到多个认证要求的约束；在这样的系统中有效地利用资源需要开发全新的调度技术。在调查这篇论文中采用的方法是首先确定当前方法的主要弱点，使认证变得繁琐。一旦了解了这些弱点，就提出了表示MC系统的新模型，以及用于量化构建这些系统的技术有效性的度量。然后对可证明系统中的资源分配和调度问题进行系统研究，目的是提供数量上优越的资源分配方法。该项目的成果将使嵌入式安全关键系统设计人员能够提供比目前更有效地利用平台资源的系统，并以更低的成本通过认证。技术转让将通过正在进行的和新的工业合作以及通过继续参与旨在改进认证过程的工业倡议来实现。结果将通过出版物、演示和教程以及在互联网上分发软件来传播。

This research addresses issues arising from the convergence of two important trends in embedded systems. One, many safety-critical applications are subject to certification requirements. Two, there is an increasing trend towards integrated architectures that support multiple functionalities, often of different criticalities, upon a single computing platform. As such systems become increasingly more complex, obtaining required certifications becomes more challenging. This project investigates the following thesis: Scheduling theory in its current form is unsuited to the design of such mixed-criticality (MC) systems that are subject to multiple certification requirements; efficient resource use in such systems requires the development of fundamentally new scheduling techniques. The methodology adopted in investigating this thesis is to first identify major weaknesses with current approaches, that render certification cumbersome. Once these weaknesses are understood, new models are proposed for representing MC systems, and metrics derived for quantifying the effectiveness of techniques for building these systems. A systematic study of resource allocation and scheduling issues in certifiable systems is then conducted, aimed at providing quantitatively superior resource allocation methodologies.The outcomes of this project will enable embedded safety-critical systems designers to provide systems that make far more efficient use of platform resources than is currently possible, and that pass certification at a significantly lower cost. Technology transfer will be achieved via ongoing and new industrial collaborations, and by continued participation in industry initiatives aimed at improving the certification process. Results will be disseminated via publications, presentations and tutorials, and distribution of software over the internet.