We present a system architecture which facilitates enhanced availability of tightly coupled distributed systems by temporarily relaxing constraint consistency.

Three different types of consistency are distinguished in tightly coupled distributed systems - replica consistency, concurrency consistency, and constraint consistency. Constraint consistency defines the correctness of the system with respect to a set of data integrity rules (application defined predicates). Traditional systems either guarantee strong constraint consistency or no constraint consistency at all. However, a class of systems exists, where data integrity can be temporarily relaxed in order to enhance availability, i.e. constraint consistency can be traded against availability. This allows for a context- and situation-specific optimum of availability.

This paper presents the basic concepts of the trading process and the proposed system architecture to enable a fine-grained tuning of the trade-off in tightly coupled distributed systems.