Fundamentals of Data Replication, Backup, and Disaster Recovery

Data replication software is traditionally used for disaster recovery (DR) scenarios such as backup, failover, and drills to ensure data safety and business continuity. The source of this article is the China Data Replication Industry White Paper (2022).

(1) Basic concepts of disaster recovery

Key metrics for evaluating a DR system are Recovery Time Objective (RTO) and Recovery Point Objective (RPO). RTO defines the maximum tolerable downtime for a business service, while RPO defines the maximum tolerable data loss.

Based on goals and cost, DR levels range from data‑level, application‑level to business‑level, with higher levels requiring greater investment.

Cloud DR (DRaaS) treats disaster recovery as a service; tenants configure backup rules on a DR management platform to protect multi‑tenant data. Compared with on‑premises DR, cloud DR faces bandwidth constraints and requires strong compression and resumable transfer capabilities.

DR providers can be cloud platforms or specialized DR vendors; the latter often offer better granularity, compatibility, cross‑platform migration, and data protection.

DR drills simulate disaster scenarios (e.g., system outage, earthquake) to test the readiness of DR systems, especially in finance, healthcare, and government.

Business Continuity Management (BCM) extends DR by establishing comprehensive response and recovery plans to improve risk mitigation and maintain operations across all enterprise functions.

(2) Backup and related knowledge

Backup is the foundation of DR, copying data from primary storage to secondary media. Backup classifications include:

A. By frequency: Scheduled (periodic) backup and real‑time backup. Real‑time backup (continuous data protection, CDP) replicates data instantly, minimizing loss.

B. By data volume: Full backup, incremental backup, and differential backup. Incremental backs up only changes since the last backup point, saving storage and time; differential backs up changes since the last full backup.

Copy Data Management (CDM) reduces redundant copies by managing data lifecycle, often using a “full‑once + permanent incremental” approach across system, database, and application layers.

Image (snapshot) backup captures block‑level data for entire volumes, enabling fast full and incremental restores, especially for large file systems.

Deduplication removes duplicate data blocks, while compression reduces data size without loss, both improving storage efficiency.

C. By backup target: Block‑level, file‑level, and database backup. Block‑level works with disk sectors, offering higher efficiency; file‑level operates on individual files; database backup handles tables, indexes, logs, etc.

(3) Disaster recovery and related knowledge

DR involves multiple geographically separated sites (e.g., two‑site three‑center architecture) to provide redundancy. DR systems must have component redundancy, sufficient distance (≈50 km), and ensure data consistency, availability, and recoverability.

DR levels:

Data‑level DR: Remote backup of data; lower cost but longer recovery time.

Application‑level DR: Synchronous or asynchronous replication of applications; higher cost, suitable for critical sectors like banking.

Business‑level DR: Full‑scale redundancy of all core services, requiring extensive infrastructure; highest cost and complexity.

High availability (HA) aims to minimize downtime from planned maintenance or unexpected failures, often implemented via active‑standby or active‑active configurations.

While HA focuses on system redundancy, DR encompasses broader switch‑over technologies (SAN/NAS, remote mirroring, IP‑based interconnects, snapshots) and generally incurs lower cost and risk than full active‑active setups.