Research Planning Grant: Adaptive Performability Management

9308912 Tai在许多实时分布式系统中，性能和可靠性都是至关重要的，影响系统本身及其环境这两个属性的条件可能会动态变化。为了确保高性能可靠的计算，计算机/通信系统的操作策略和要求必须作出相应的调整。本研究研究了适应性绩效管理（绩效是性能和可靠性的综合度量），通过基于运行时模型的决策来同时保证性能和可靠性目标。该方法将为计算机系统提供基于底层可执行性模型来确定其是否达到其性能和可靠性目标的能力。否则，它将选择另一种策略。本研究的目的是研究对象系统运行时可执行性概念和建模技术的效用，并研究由此产生的方法在确保高性能可靠计算方面的有效性。因此，重点将放在：1)强调性能，可靠性和其他关键属性之间的权衡并促进自适应可执行性管理的模型构建方法，2)模型解决方法足够有效以支持运行时决策，以及3)评估自适应可执行性管理方案有效性的方法。＊＊＊

9308912 Tai In many real-time distributed systems where both performance and dependability are critical, the conditions affecting these two attributes in the system itself as well as its environment may change dynamically. In order to assure high performance dependable computing, a computer/communication system's operational strategies and requirements have to be adjusted responsively. The research studies adaptive performability management (performability is a combined measure of performance and dependability) for assuring performance and dependability goals simultaneously via run-time model-based decision making. The method will provide a computer system with the capability to determine, based on the underlying performability model, whether it achieves its performance and dependability goals. If not, it will select an alternative strategy. The objectives of this research are to investigate the utility of performability concepts and modeling techniques when the object system is in operation, and to investigate the effectiveness of the resulting methodology in assuring high performance dependable computing. Accordingly, the focus will be on: 1) methods of model construction which stress the tradeoffs among performance, dependability and other critical attributes and facilitate adaptive performability management, 2) methods of model solution which are sufficiently efficient to support run-time decision making, and 3) methods for evaluating the effectiveness of an adaptive performability management scheme. ***