A Control-Theoretical Approach to Performance Guarantees in Performance-Critical Systems

许多重要的应用程序必须提供有保证的实时性能，尽管工作负载不确定，任务的计算时间变化很大，并且有大量的交互站点。这类应用的例子包括智能空间、互联网上的金融市场、支持敏捷制造的工厂集合以及高科技战场协作。这项研究的目标是开发一种面向软件的反馈控制理论和实践，为这些类型的系统提供综合性能保证。研究了在这些类型的应用的操作系统中嵌入这种控制器的方法。这项研究的最终愿景是，软件设计人员将能够对软件系统的各部分进行建模，并使用这些模型来开发基于反馈控制理论的软件控制算法。这项工作为设计和分析在大量不确定性下运行的大型系统的总体行为奠定了科学基础。这些解决方案可以嵌入到操作系统中，以满足瞬变和稳定状态下的性能规格，如截止期错失率、稳定性、超调、稳定时间和灵敏度要求。

Many important applications must provide guaranteed real-time performance in spite of uncertain workloads, highly varying computation times for tasks, and with a large number of interacting sites. Examples of such applications include smart spaces, financial markets on the Internet, collections of factories supporting agile manufacturing, and high-tech battlefield coordination. The objective of this research is to develop a software-oriented theory and practice of feedback control that will provide aggregate performance guarantees for these types of systems. Methods to embed such controllers in the operating systems of these types of applications is investigated. The ultimate vision of this research is that software designers will be able to model parts of software systems and use those models to develop software control algorithms based on a theory of feedback control.This work establishes a scientific basis upon which to design and analyze the aggregate behavior of large systems that operate under a great deal of uncertainty. The solutions can be embedded in operating systems to meet the performance specs in transient and steady states, such as deadline miss ratio, stability, overshoot, settling time, and sensitivity requirements.