Maintaining consistency of data in mobile distributed environments

To deal with the frequent, foreseeable and variable disconnections that occur in a mobile environment, we introduce a flexible, two-level consistency model. Semantically related or closely located data are grouped together to form a cluster. While all data inside a cluster are mutually consistent, degrees of inconsistency are allowed among data at different clusters. To take advantage of the predictability of disconnections, and to accommodate mobility, the cluster configuration is dynamic. We allow transactions to exhibit certain degrees of tolerance for inconsistencies by introducing strict and weak operations. Weak operations are operations that can be executed under weaker consistency requirements. We define correctness criteria for schedules that involve weak operations and compare them with traditional serializability criteria. Finally, we argue that our model is appropriate for a variety of other environments including very large databases and multidatabases.