Basser Seminar Series

Consistency: the never-mentioned flaw in the SOA vision

Paul Greenfield
Research Scientist
CSIRO Mathematical and Information Sciences

Wednesday 28 November 2007, 4-5 pm

School of IT Building, Lecture Theatre, Room 123, Level 1


The SOA vision of distributed service-based systems is alluring but fundamental deficiencies in the underlying technologies mean that these Internet-scale distributed applications will prove fragile and hard to build in practice. The problems come from the need to ensure the consistency of the participating applications in the face of errors and failures. The traditional solution here has been to use distributed transaction technologies but these are not applicable in the loosely-coupled and untrusting world of open services running over the open Internet.

This talk describes the results of a multi-year research collaboration between CSIRO and the University of Sydney into this question and the solutions we developed to provide the equivalents of atomicity and isolation. The approach to the 'atomicity' problem was to see it as effectively a protocol correctness question and use formal methods techniques to show that correctly implemented applications could not finish in globally-inconsistent states. We also developed a run-time protocol that could detect consistency without needing a global coordinator. The problems caused by the lack of isolation are addressed by our work on Promises, an approach that uses distributed predicates rather than locks to avoid conflicts over shared resources without unnecessarily blocking concurrent applications.

Speaker's biography

Paul Greenfield is a research scientist in CSIRO Mathematical and Information Sciences, and an Honorary Associate of the School of IT at the University of Sydney. He was once a lecturer in the Basser Department of Computer Science before escaping to work in industry doing software R&D. He moved to CSIRO in 1997 and was an author on 25 papers from his work there on distributed computing. He is now working on bioinformatics and genetics applications of database technologies.