Building Blocks for High Performance, Fault-Tolerant Distributed Systems

摘要

This project developed theory and systems support to aid in the construction of adaptive, survivable distributed systems. The systems are designed to run in highly dynamic environments such as the internet, wireless networks, and sensor networks. Participating processes may join, leave, and fail during computation. The systems that were considered solve problems of data sharing and management, resource sharing and management, communication, and coordination. Specifically, the project involved developing reusable 'building blocks' global service specifications and distributed algorithms-for dynamic distributed systems. The work included an extensive study of view- oriented group communication services and algorithms, which is now 'transitioning' into use at Lincoln Laboratories. A major focus was on design and analysis of algorithms for implementing reliable atomic shared memory in highly dynamic networks. Other algorithmic work covered dynamic algorithms for atomic broadcast, scalable reliable multicast, and topology control. In addition, the project produced results on mathematical semantic foundations to support modeling and analysis of highly dynamic distributed systems, and on tools to support this effort.