Processor Specialization in Fault-Tolerant Distributed Systems

该项目的目标是开发一个可伸缩的容错框架，该框架可以定制，以经济地覆盖广泛的故障模型、应用程序和计算环境。该系统基于n模冗余（NMR）概念，统一处理所有类型的故障。为了提高系统性能并增强其可伸缩性，开发了一种能够容忍恶意故障的专用NMR （SNMR）协议。在这项研究中，SNMR协议将在几个方面得到增强，包括适应广域网应用和有效地容忍一般和故障停止故障。通过调整复制级别、纠错算法等参数，可以定制系统，满足不同的应用需求和成本约束。SNMR核心的原型将在两年期间实施。将深入研究SNMR系统的可定制性、适用性、性能、故障覆盖和可扩展性，并与现有方法进行比较。＊＊＊

The goal of this project is to develop a scaleable fault tolerance framework that can be customized to economically cover a wide range of fault models, applications, and computational environments. The system is based on the N-modular redundancy (NMR) concept to uniformly handle all types of failures. A specialized NMR (SNMR) protocol that can tolerate malicious failures has been developed to improve the system performance and to enhance its scalability. In this research, the SNMR protocol will be enhanced in several ways, including adaptations for wide- area network applications and for efficiently tolerating general and fail-stop failures. By adjusting the parameters, such as the replication level and error-checking algorithms, the system can be customized to satisfy various application requirements and cost constraints. A prototype of the SNMR core will be implemented in the two-year period. The customizability, applicability, performance, fault coverage, and scalability of the SNMR system will be thoroughly studied and compared with existing approaches. ***