Robust Real-Time System Resource Management Using Feedback

实时计算是当前和未来许多应用领域的一项使能技术。许多未来一代的实时系统被期望是高度动态的，并在严格的时间约束下运行在容易发生故障的确定性环境中。因此，这些系统需要在提供高性能的同时具有健壮性。这激发了对健壮的资源管理技术的需求，这些技术能够动态地满足实时需求，并在存在不确定性的情况下提供优雅的降级。尽管实时系统在资源管理方面取得了大量成果，但大多数成果都是基于“开环”策略的，当能够准确地对工作负载建模时，这些策略是有效的。这些方案不足以解决工作负载无法准确建模的许多实际问题。因此，需要一种有效的资源管理体系结构，以便在存在不确定性的情况下能够获得可预测的性能保证。反馈控制理论一直是建模在不确定环境中运行的系统的核心。在过去的几十年里，这一理论在这个方向上取得了令人印象深刻的进展。反馈控制理论表明，正确的自适应将在稳健性方面产生显著的红利。本研究致力于为实时系统开发一种采用反馈控制策略的健壮资源管理框架。

Real-time computing is an enabling technology for many current and future application areas. Many future generations real-time systems are expected to be highly dynamic and operate in fault-prone on-deterministic environments under strict timing constraints. Therefore, these systems need to be robust while delivering high performance. This motivates the need for robust resource management techniques that dynamically address real-time requirements and provide graceful degradation in the presence of uncertainty. Despite the significant body of results in resource management in real-time systems, most of them are based on``open-loop'' strategies which are effective when the workload can be accurately modeled. These schemes are inadequate for many real world problems wherein the workload cannot be accurately modeled. Thus, there is a need for efficient architectures for resource management where predictable performance guaranteescan be obtained in the presence of uncertainty.Feedback control theory has been central to modeling systems operating in uncertain environments. In the past few decades, this theory has made impressive strides in this direction. Correct adaptation as illustrated by feedback control theory will yield significant dividends with respect to robustness. This research focuses on developing a robust resource management framework for real-time systems employing feedback controlstrategies.