CAREER: Composition Approaches for the Analysis of Complex Systems

技术进步正在产生越来越复杂的系统，这些系统通常被期望或要求满足性能要求，但又极其可靠和安全。因此，工程师们越来越多地在设计过程中利用计算机工具来分析系统的高级模型。设备故障或请求到达等不确定性会导致模型中的随机事件。自动化计算机工具可以对模型所描述的随机过程进行数值分析以获得系统模型的性能测量，并且可以利用隐式数据结构来表示比否则可能的更大的随机过程。或者，离散事件模拟可以用来产生绩效测量的可信区间。这个NSF职业研究过程项目通过概括隐式表示支持的模型类型来扩展高级系统模型的适用性，以包括一个统一的层次结构，其中模型事件本身可以由另一个更详细的模型来描述。这使得复杂的系统模型可以由几个简单的组件构成。其目标是使用隐式方法来控制困扰复杂系统分析的状态空间爆炸。该项目正在软件库和工具中开发和实现构建和利用模型层次结构的技术。这些技术包括使用隐式表示法的数值分析方法、离散事件模拟，以及将数值分析方法与离散事件模拟相结合的混合方法。该项目正在开发能够分析复杂系统并影响计算机系统研究的重要领域的工具和技术，包括复杂实时嵌入式系统、对等网络、安全关键系统、基于代理的系统以及高置信度系统的软件验证。职业研究计划通过引入离散事件模拟和随机过程分析课程的作业、项目和讲座与教育工作紧密结合。

Technology advancements are producing increasingly complex systems, that are often expected or required to meet performance demands and yet be extremely reliable and secure. As a result, engineers are increasingly utilizing computer tools to analyze high-level models of a system during its design. Uncertainties, such as device failures or arrivals of requests, lead to stochastic events within the model. Automated computer tools can numerically analyze the stochastic process described by the model to obtain performance measures of the system model, and can utilize implicit data structures to represent much larger stochastic processes than would be possible otherwise. Alternatively, discrete-event simulation can be used to produce confidence intervals for the performance measures.This NSF CAREER research process project extends the applicability of high-level system models by generalizing the types of models supported by implicit representations to include a unified hierarchy in which a model event may itself be described by another more detailed model. This allows a complex system model to be constructed out of several simple components. The goal is to use implicit methods to control the state space explosion that has plagued analysis of complex systems. Techniques to construct and exploit the model hierarchy are being developed and implemented by this project in software libraries and tools. These techniques include numerical analysis methods that use implicit representations, discrete-event simulations, and hybrid approaches that unify numerical analysis methods with discrete-event simulation. The project is developing tools and technology that will enable the analysis of complex systems and impact important areas of computer systems research, including complex real-time embedded systems, Peer-to-Peer networks, safety-critical systems, agent-based systems, and software verification for high-confidence systems.The CAREER research program is closely integrated with educational efforts through assignments, projects, and lectures that are introduced into courses on discrete-event simulation and analysis of stochastic processes.