Expressing database transactions as atomic-operations

DBMS are widely used and successfully applied to a huge range of applications. However the transaction paradigm that is, with variations like nested transactions and workflow systems, the basis for durability and correctness is poorly suited to many modern applications. Transactions do not scale well and behave poorly at high concurrency levels and in distributed systems. The atomic-operations model, which is proposed in this paper, is an alternative database paradigm based upon a complete isolation between application and database and on explicit consistency management. We describe this model and outline classes of applications for which atomic-operations might provide advantages such as better performance than transactions. We describe a graph-based approach that expresses database transactions consisting of four types of graph (operation, consistency, combined and atomic) and a transformation algebra (the operators ⊕, ⊗, ·, Φ and Ψ) to enable transactional applications to be rewritten into forms usable under the atomic-operations model.