What Are the Five Core Data Replication Techniques for Disaster Recovery?

The 2021 whitepaper, compiled over ten months by dozens of frontline engineers and industry experts, provides a comprehensive overview of disaster‑recovery (DR) technologies, focusing on data replication methods that underpin modern backup and business‑continuity strategies.

1. Application‑and‑Middleware Level Replication

Application‑level replication synchronizes or asynchronously writes data between primary and standby databases via the application itself, ensuring data consistency while allowing the DR site to run concurrently with production. This method offers high customizability and direct control but introduces complexity, tight coupling with business logic, and increased risk of data loss.

2. Host‑Level Replication

Host‑level replication creates disk‑volume mirrors or copies, operating at the host’s volume manager layer. It is hardware‑agnostic, provides reliable high‑throughput transfer over IP networks, and enables rapid recovery from a failed primary site. Drawbacks include CPU overhead on the host and limited snapshot capabilities.

The article highlights a byte‑level incremental asynchronous replication technology developed by a UK vendor, which captures file‑system write operations (e.g., rename, set‑attr) via a bypass engine and transfers serialized data using a Data‑Order Transfer (DOT) protocol, minimizing CPU and storage I/O impact.

3. Database‑Level Replication

Database replication includes physical (block‑level) and logical (SQL‑based) methods. Logical replication streams redo and archive logs to a standby site, offering asynchronous updates without guaranteeing real‑time consistency. Physical replication copies redo logs directly, supporting read‑only standby instances. An open‑platform solution parses online or archived logs to apply row‑level changes, enabling multi‑site active‑active configurations.

4. Storage‑Gateway Replication

Storage gateways sit between servers and storage arrays, providing virtualization, remote replication, heterogeneous storage integration, and continuous data protection. By offloading replication workloads from servers, they support diverse platforms and improve manageability in SAN environments, though performance depends on bandwidth and gateway capabilities.

5. Storage‑Media Replication

Media‑based replication uses built‑in firmware, operating‑system features, IP networks, or Fibre Channel to copy data synchronously or asynchronously between storage systems. It supports one‑to‑one, one‑to‑many, or many‑to‑one topologies and can operate at block level, often requiring identical storage models and low‑latency, high‑bandwidth links.

The whitepaper also compares these techniques in terms of data volume transferred, showing a typical 300 MB file with a 512 B change results in transfer ratios of 4800:128:1 for traditional, host‑level, and byte‑level replication respectively.

Finally, replication modes are classified as continuous (real‑time) versus point‑in‑time copies. Continuous replication includes synchronous (writes must complete on both sides) and asynchronous (writes proceed without waiting) variants, while point‑in‑time copies protect against logical errors by capturing the system state at specific moments.