The paper presents a "majority consensus" algorithm for maintaining synchronization in multiple copy databases, addressing the update synchronization problem. The algorithm is designed to function effectively even in the presence of communication and database site outages. It ensures both mutual consistency (all copies converge to the same state) and internal consistency (preservation of invariant relations within the database). The algorithm uses distributed control, where each database site votes on update requests, and only a majority approval is required for a request to be accepted. The voting and request resolution rules prevent conflicting updates from being accepted simultaneously, ensuring mutual consistency. The update application rule guarantees that recent updates supersede older ones, maintaining internal consistency. The algorithm's correctness is demonstrated through arguments that show it is deadlock-free, converges all copies to the same value, and implements mutual exclusion for accepted updates. The paper also discusses the robustness of the algorithm, its ability to recover from failures, and its performance characteristics.The paper presents a "majority consensus" algorithm for maintaining synchronization in multiple copy databases, addressing the update synchronization problem. The algorithm is designed to function effectively even in the presence of communication and database site outages. It ensures both mutual consistency (all copies converge to the same state) and internal consistency (preservation of invariant relations within the database). The algorithm uses distributed control, where each database site votes on update requests, and only a majority approval is required for a request to be accepted. The voting and request resolution rules prevent conflicting updates from being accepted simultaneously, ensuring mutual consistency. The update application rule guarantees that recent updates supersede older ones, maintaining internal consistency. The algorithm's correctness is demonstrated through arguments that show it is deadlock-free, converges all copies to the same value, and implements mutual exclusion for accepted updates. The paper also discusses the robustness of the algorithm, its ability to recover from failures, and its performance characteristics.