Overview of Data Replication Technologies for Disaster Recovery

The 2021 China Disaster Recovery Whitepaper, compiled over ten months by dozens of engineers and experts, provides a comprehensive knowledge base on disaster recovery technologies, including the latest products and solutions for backup, data management, and cloud disaster recovery.

Data replication (Replication) copies data across locations in real‑time or on a schedule, and is classified into five layers:

1. Application and middleware layer: Replicates data via application‑level read/write operations, offering customizable, transparent replication but with high implementation complexity and increased risk of data loss.

2. Host layer: Uses disk‑volume mirroring or copying, independent of underlying storage, providing flexibility but consuming CPU resources and requiring advanced software support.

3. Database layer: Includes physical and logical replication; logical replication uses redo/archived logs for asynchronous copy, while physical replication copies redo logs for synchronous or asynchronous persistence, supporting read‑only replicas and multi‑active databases.

4. Storage‑gateway layer: Positioned between servers and storage, leveraging storage virtualization to offer remote replication, heterogeneous storage integration, high‑availability imaging, snapshots, and continuous data protection.

5. Storage‑media layer: Employs built‑in firmware or OS mechanisms over IP or Fibre Channel to perform synchronous or asynchronous block‑level copying, supporting one‑to‑one, one‑to‑many, or many‑to‑one replication topologies.

The whitepaper also highlights byte‑level incremental asynchronous replication, which captures write operations at the OS memory level, incurs negligible CPU load, avoids storage I/O, and works efficiently over low‑bandwidth links, making it suitable for hybrid and cloud environments.

Specific tools such as i2Active, an Oracle real‑time replication solution based on redo‑log analysis, are described for low‑latency, non‑intrusive data migration and disaster backup.

A comparative table and graphs illustrate the data volume transferred by different replication methods, showing the efficiency of byte‑level and variable‑block techniques.

Replication modes are further divided into continuous replication—providing ongoing data sync for rapid recovery during failures—and point‑in‑time copy—capturing data snapshots for restoration after logical errors or malware incidents.