Search

This report summarizes the viewpoints and insights gathered in the Dagstuhl Seminar on Atomicity in System Design and Execution, which was attended by 32 people from four different scientific communities: database and transaction processing systems, fault tolerance and dependable systems, formal methods for system design and correctness reasoning, and hardware architecture and programming languages. Each community presents its position in interpreting the notion of atomicity and the existing state of the art, and each community identifies scientific challenges that should be addressed in future work. In addition, the report discusses common themes across communities and strategic research problems that require multiple communities to team up for a viable solution. The general theme of how to specify, implement, compose, and reason about extended and relaxed notions of atomicity is viewed as a key piece in coping with the pressing issue of building and maintaining highly dependable systems that comprise many components with complex interaction patterns.

The concept of coordinated atomic (CA) actions was introduced about two years ago. Since then this research has addressed many new problems and, in particular, problems of CA action distribution and implementation. Recently we have gained some experience in implementing different schemes in Ada 95 and Java. Our intention within this paper is to discuss how distributed CA action schemes can be realised. In particular, we outline different ways of action component distribution, trade-offs, applications for which these schemes are applicable. We discuss a wide range of schemes (some of them have not yet been implemented) based on a classification of various approaches to CA action distribution; to do this we analyse all possible ways of different action component distribution. We believe that this general discussion should help to better understand the current state of CA action implementation and is important for future research in CA actions.

The purpose of this paper is to propose a way of tolerating software (design) faults in distributed systems relying on the well-known conversation (atomic action) approach. To do this, we shall consider differences between two programming paradigms: group communication and conversations, and discuss how a group communication service can be used to provide design fault tolerance by conversations. The main characteristics and peculiarities of this new conversational group service are described.

Atomic actions have been proposed in the literature [Lomet, 77] as mechanisms for process structuring. They have the property that either the atomic action successfully executes or if it aborts it is as if it never executed. Thus they combine features of synchronization and error recovery. Atomic actions are of particular interest in distributed systems . In a distributed system there may exist individual object managers guarding a group of objects and transactions are composed using requests to these decentralized managers. In such organizations, conventional synchronization mechanisms like monitors and other synchronizers fail to provide robust synchronization. In addition, they cannot support post-hoc interfaces and require advance knowledge of all potential sharing [Reed, 79].