Enterprise organizations running mission-critical Management Information Systems (MIS) face a modernization paradox that has no clean answer in the existing literature. The systems most urgently needing replacement are precisely those that cannot be taken offline. Traditional migration strategies handle application concerns reasonably well but leave a serious gap at the data layer, where both legacy and cloud systems must simultaneously read and write consistent information for years at a time. This paper presents the Live Data Coexistence Framework (LDCF), a validated five-layer bidirectional architecture for zero-downtime enterprise MIS modernization. The framework addresses the full scope of the coexistence problem: a physical replication pipeline supporting heterogeneous sources including mainframe z/OS and relational databases; an eight-pattern integrity and correctness framework; a metadata-driven transformation engine that removes the schema-parity constraint built into all standard CDC tools; a Kafka-based reverse synchronization path for cloud-to-legacy data flow; and an automated self-healing reconciliation layer that bounds drift to a twenty-four-hour detection window. A configuration-driven table lifecycle mechanism allows individual tables to be retired from synchronization incrementally as their migration completes, with no code change or pipeline restart required, giving teams a structured and reversible path from full coexistence toward full cloud operation. Eight original engineering patterns are introduced, among them the Binary Key Bridging Pattern, which is the first documented systematic solution for handling mainframe hardware-clock-generated binary surrogate keys in CDC pipelines, and a Metadata-Driven Transformation Engine enabling 1-to-many table decomposition during active replication. The framework has been validated in a production insurance management environment, successfully migrating more than three hundred legacy tables, creating over fifty cloud-native tables through schema decomposition, processing more than one million DML transactions, and sustaining approximately 1.2 million file operations weekly, with a failure rate below 0.1 percent and zero unrecovered data loss across an extended period of continuous operation.
Krishan Kumar (Sat,) studied this question.