Research Initiation Award: An Integrated Framework for Fault-Tolerant Applications in Real-Time and Non-Real-Time Systems

本文研究了一种新的计算模型，该模型扩展了具有时间约束和资源需求规范的面向对象模型。定义了一组场景（一次执行的一组应用程序）和一种场景更改机制，该机制允许根据情况执行不同的应用程序集。这些场景更改由错误处理程序、操作员或用户触发。场景更改还允许多种类型容错行为的一致性，例如优雅降级、负载脱落等。研究了这些场景变化的延迟，并量化了它们在实时和非实时系统中的适用性。开发了用于检测可应用这些场景变化的情况的方法，并开发了将这些场景变化适应于无时间约束的应用和系统的理论。更准确的模型是基于实际工作负载和情况开发的，以模拟实际故障模型，包括瞬态故障和永久故障。这些是暂时的或物理性质的（例如，瞬态过载，站点崩溃或链接故障）。下面是支持这种计算模型的容错实时系统原型的实现。

This research studies a new model of computation, which extends the object-oriented model with timing constraints and resource requirements specifications. A set of scenarios (a set of applications executing at a time) and a scenario change mechanism that allows for different sets of applications to be executed according to the situation are defined. These scenario changes are triggered by an error handler, by an operator or by a user. Scenario changes also allow uniformity in multiple types of fault-tolerant behavior such as graceful degradation, load-shedding, etc.. The latency of these scenario changes and quantify their applicability to real-time and non-real-time systems are studied. Methods are developed for the detection of situations in which these scenario changes can be applied, and develop a theory for the adaptation of these scenario changes to non-time-constrained applications and systems. More accurate models are developed based on real-life workloads and situations, to model realistic fault models, encompassing transient as well as permanent faults. These are of temporal or physical nature (e.g., transient overloads, site crashes or link failures). The implementation of a prototype of a fault-tolerant real-time system that supports this model of computation follows.