RIA: The Application Oriented Fault Tolerance Paradigm for Multicomputer Systems

本项目致力于识别适用于一种新的容错设计方法--面向应用范型的问题类。在该方法中，为应用程序程序员提供了一个可靠的并行处理模型，该模型在功能上由用于约束和检测故障行为的应用层约束谓词、用于在出现故障组件时操作的配置控制层和分布式诊断基础组成。应用程序程序员必须只关心生成每个单独程序的显著可靠性特性的抽象。该任务与从规范阶段开始的软件生命周期的每个阶段协同执行。该方法被应用于大规模并行多计算机网络的并行软件设计，该软件可以在存在硬件和软件错误的情况下运行。解决的问题包括系统问题和配置控制、应用程序适用性、自动约束谓词生成以及故障检测和恢复。在大型分布式多计算机系统中，容错是至关重要的。在同时存在硬件和软件错误的大规模并行多计算机网络中，设计一种系统的方法来开发可靠的控制系统是非常重要的。本研究发展了一种系统的方法，并证明了其在大规模并行系统设计中的适用性。强烈建议您提供支持

This project focuses on identifying problem classes applicable to the Application Oriented Paradigm, a new class of fault-tolerance design method. In this approach, the applications programmer is presented with a Reliable Parallel Processing Model, which functionally consists of an application level constraint predicate for constraining and detecting faulty behavior, a configuration control layer for operation in the presence of failed components, and a distributed diagnostic basis. The applications programmer must only be concerned with the generation of an abstraction of each individual program's salient reliability features. This task is performed in concert with each phase of the Software Life Cycle beginning with the specification phase. The methodology is applied to the design of parallel software for massively parallel multicomputer networks, which can function in the presence of both hardware and software errors. Issues addressed include system issues and configuration control, application applicability, automated constraint predicate generation, and fault detection and recovery. Fault tolerance is of paramount importance in large distributed multicomputers. The design of a systematic method for developing reliable control systems in massively parallel multicomputer networks in the presence of both hardware and software error is very important. This research develops a systematic method and demonstrates its applicability on applications in the design of massively parallel systems. Support is strongly recommended