Abstract: This research has developed an approach for analyzing data dependencies in a distributed environment, providing a rule-based mechanism to support semantic correctness in the recovery of concurrently executing processes over Grid Services. Delta-Enabled Grid Services are used to capture incremental data changes, known as deltas, from processes that execute over distributed services. Deltas are forwarded to a Process History Capture System (PHCS) that constructs a global process execution history to support the analysis of data dependencies when process failure occurs. An abstract execution model has been developed that is composed of three sub-models: (1) a service composition and recovery model defining the hierarchical composition structure with recovery features for backward recovery and forward execution; (2) a process dependency model defining and analyzing read and write dependencies among concurrently executing processes; and, (3) a rule-based model that uses process interference rules to specify how failure recovery of one process can potentially affect other process execution based on application semantics. A Process Recovery System (PRS) implements the recovery algorithms associated with the abstract execution model. A simulation framework has been developed to demonstrate the functionality of the PHCS and PRS for concurrent process recovery and to conduct performance evaluation on the PHCS and PRS. The results of this research support relaxed isolation and application-dependent semantic correctness for concurrent process execution, with a unique approach to resolving the impact of process failure and recovery on other concurrently executing processes, using data dependencies derived from distributed, autonomous services.