A middleware service for fault-tolerant group communications

Abstract

Many distributed applications require multicast group communication services, enabling an entity to interact with a group of other entities. Providing the reliability and ordering guarantees required by group based applications is not a trivial task in distributed systems where computation and communication delays might not be known accurately. Furthermore, the approaches available to support these guarantees are diverse. The choice of approach may significantly effect the performance of an application and/or may not be suitable for some application types. Nowadays, distributed applications are frequently built as a Middleware service. The Thesis develops techniques for providing group communication support in Middleware environments. A group communication service has been designed and implemented in such a way as not to hinder the interoperability/portability of applications built using it. The service provides a variety of functions that may be tailored to suit many different types of applications. Group communication protocols are presented that ensure reliability and ordering guarantees. Furthermore, the reliability and ordering guarantees of such protocols may be tailored to suit a wide variety of applications. Mechanisms that provide a variety of approaches to inter-member and inter-group interactions that are suitable for satisfying the requirements of many different types of applications (e.g., fault- tolerant, collaborative) are also supported. The service can work over local and wide area networks (Internet).