| In recent years, the study of distributed systems has become an increasingly important focus of computer science research. One class of distributed system that has become widely used is the dependable, or fault-tolerant distributed system. Building a dependable distributed system requires addressing a number of important issues, such as communication reliability, message ordering, quality of service, shared state maintenance, and failure detection and recovery. Central to all of these concerns is the issue of scalability, which is becoming increasingly important as the focus of distributed system development moves toward large-scale, wide-area, heterogeneous systems.; One popular paradigm which has been developed for structuring dependable distributed systems is the group communication model. Unfortunately, experience has shown that group communication often does not scale well for large systems. Scalability issues typically manifest themselves in two ways: decreasing performance as the system grows larger, and great increases in the time and effort required to design and implement the system. Consequently, our research has focused on two complementary areas: enabling predictable and timely group communication, group communication, we can alleviate the performance impact of increasing system size and complexity while still taking advantage of the message delivery and fault-tolerance guarantees provided by process groups. By developing new design and construction by bridging the gap between the low-level services typically offered by group communication middleware and the high-level needs of the applications that use it. |
