Incremental Data Integration for SQL and NoSQL Stores with Conflict Resolution

Abstract

Incremental integration across SQL and NoSQL stores is difficult because relational rows and document-based records differ in schema structure, update granularity, identity representation, and consistency behavior. This article presents a conflict-aware integration framework that combines incremental change extraction, canonical record modeling, version-vector ordering, schema mapping, conflict detection, policy-based resolution, and reconciliation-window validation. The framework is designed to synchronize row-level SQL changes and field-level NoSQL document changes without relying on unsafe last-write-wins logic. Simulated results show that insert-duplicate and schema-shape conflicts achieved the highest resolution success rates of 96.8% and 94.1%, while updatedelete conflicts remained the most difficult at 78.2%. Reconciliationwindow analysis showed that integration accuracy increased from 91.2% at a 1- minute window to 98.9% at a 60-minute window, while pending conflict count decreased from 148 to 28. These findings show that reliable SQL-NoSQL integration requires conflict-specific policies, version-aware ordering, and delayed consistency confirmation through reconciliation windows.