University of Houston
Department of Computer Science

In partial fulfillment of the Requirements for the Degree of
Doctor of Philosophy

José Carlos Martínez
will present his dissertation defense

A Majority Consensus Implementation of Fault-Tolerant CORBA

Abstract

The proposed standard for Fault-Tolerant CORBA envisions the management of replicated data within a group communication framework. We propose an alternative majority consensus approach that does not require any knowledge of the network topology and guarantees the consistency of the replicated data in the presence of any non-Byzantine failure. We take advantage of CORBA’s support of interoperability, portability and location transparency to provide transparent management of the replicas to the client applications. Our protocol relies on dynamic voting and volatile witnesses to reduce its storage and message overhead. Each witness only contains a bitmap describing which entities belong to the current majority block (the cohort set). This cohort set is kept in main memory, which lowers even further their cost while reducing their response time by a factor of three. We support the idea of upgradeable witnesses. Witnesses should be able to become replicas in order to maintain the correct balance within both entity types (replicas and witnesses). The protocol is proposed as a suitable solution to fulfill the strong fault tolerant requirements of mission-critical applications. We found that three replicas and two witnesses managed by our protocol had similar availability and average response time as the best group communication protocols. Our protocol clearly outperforms object group communication for update requests with the same number of replica entities.

We also introduce an architecture pattern for the deployment of any fault tolerant strategy as well as the containerization of the framework for open, web-enabled, component-based enterprise applications.

Date: Wednesday, May 16, 2001
Time: 3:00 PM
Place: 550-PGH

Faculty, students, and the general public are invited.
Dissertation Advisor: Dr. Jehan-François Pâris