This work studies the mechanisms of reliable broadcast communications that, as a natural extension of traditional point-to-point services, try to provide the adequate support for the implementation of distributed applications.
The properties of broadcast protocols are analyzed from different points of view, including group management and addressing, reliability and fault-tolerance, synchronism (and how it relates to application synchronism, specially in real-time environments), and how ordering disciplines are enforced on system messages. Performance and quality of service provided are evaluated in relation with the assumptions made on the tolerated faults and on the underlying communication support.
The xAMp is presented. It consists of a set of services to support the development of distributed applications, with different dependability, functionality, and performance requirements. The xAMp includes group membership, reliable multicast and time services, which are highly versatile and designed to be used in collapsed-layer architectures.
The xAMp implementation consists of a highly integrated package which exploits broadcast local-area networks and the use of fail-controlled components to provide the best tradeoff between functionality and performance.
Not available in Postscript format.