A Dynamic Light-Weight Group Service

Luís Rodrigues, Katherine Guo, Paulo Veríssimo, Kenneth P. Birman

In in Journal of Parallel and Distributed Computing Vol. 60, No. 12, December 2000, pp 1449-1479. Academic Press.

Abstract

The virtual synchrony model for group communication is a powerful paradigm for building distributed applications. Implementations of virtual synchrony usually use failure detectors and failure recovery protocols. In applications that require a large number of groups, significant performance gains can be attained if these groups share the resources required to provide virtual synchrony. A service that maps multiple user groups onto a small number of instances of a virtually synchronous implementation is called a Light-Weight Group Service.

This paper describes a new design for the Light-Weight Group protocols that enables such service to function transparently. We discuss how these protocols can be applied in dynamic environments, where group mappings cannot be defined a priori and may change over time. We show that it is possible to establish mappings that promote resource sharing and, at the same time, minimize interference. These mappings can be established in an automated manner, using heuristics applied locally at each node. Experiments using an implementation in the Horus system show that significant performance improvements can be achieved with this approach.

Luís Rodrigues