Cooperative Transactions: A Data-Driven Approach

In this paper we present a new cooperative transaction model. The model has been developed using a data-driven approach, where we have not considered constraints specific to one application class, but only data-specific constraints. This approach is more general than the application-driven approach adopted in other cooperative transaction models. The notion of set of complementary objects is intrinsic here. Two objects belong to a set of complementary objects if they are connected by dependency relations, directly or transitively. A set of users, represented by cooperative transactions within a cooperative unit, shares the consistency preservation of a set of complementary objects (global consistency). Responsibilities are separated: the set of complementary objects is divided into disjointed subsets, each being manipulated by at most one user. Global consistency is achieved by selective cooperation between users in order to take into account the dependency relations between the respective objects that they manipulate. The model enforces two grades of selective cooperation. In weak cooperation, a user makes his final effects visible to another (one-way cooperation). In strong cooperation users make their intermediate effects symmetrically visible (two-way cooperation). Global consistency is ensured when there are concurrent transactions external to the cooperative unit, and in case of failure.