Database High‑Availability Techniques: HADR, HACMP, Replication, Storage DR, and DPF

Database high availability is a complex systems engineering discipline; this article introduces several fundamental techniques—HADR, HACMP, data replication, storage‑layer disaster recovery, and DPF high‑availability—describing their technical characteristics, applicable scenarios, and how they can be combined to achieve end‑to‑end resilience from storage to application.

1. DB2 HADR – High Availability Disaster Recovery is IBM DB2’s database‑level data‑copy mechanism, originally from Informix. It uses a primary‑standby pair; before DB2 9.7 the standby was read‑only, after 9.7 it can be read‑only to offload the primary. Synchronous transmission is recommended when bandwidth is stable; asynchronous or super‑asynchronous modes are used for long‑distance links. HADR cannot be used with DPF, does not support compression, encryption, or heterogeneous replication, but integrates third‑party SSH for secure transfer.

2. SQL Replication and Q Replication – SQL replication works best within the same LAN, while Q replication (leveraging WebSphere MQ) is more tolerant of poor networks and is often paired with HADR for remote disaster recovery (e.g., China Tobacco’s DR site). Q replication reads transaction logs, has low performance impact, supports table‑level replication, and is well‑supported on DB2; Oracle’s counterpart is GoldenGate, and IBM’s CDC (formerly Data Mirror) offers multi‑database support.

3. HACMP – In Cascading mode HACMP provides a primary‑standby pair with node priority; the higher‑priority node takes over resources on failure. Rotating mode uses equal‑priority nodes, suitable for telecom services requiring high availability. Concurrent mode runs resources on all nodes without a designated primary, often combined with Oracle RAC or parallel servers. HACMP alone does not protect the database content, so it is usually paired with HADR for full protection.

4. DPF High‑Availability – DPF itself lacks built‑in HA; however, multi‑node configurations can provide limited disaster tolerance if catalog nodes remain up and non‑critical nodes fail. True HA for DPF requires external mechanisms (e.g., HACMP for OS/network failures) and careful planning of critical node backups and table‑space distribution.

5. Storage‑Layer Disaster Recovery – Technologies such as disk mirroring, SRDF, and Veritas BMR provide site‑to‑site replication at the storage level. SRDF supports synchronous, near‑synchronous, and asynchronous modes over distances of several kilometers to thousands of kilometers and works with any host or database system. Veritas BMR enables OS‑level backup and rapid system restore, with agents for online hot backups of various databases.

6. Network, Power, and Procedural Aspects – High‑availability networks require redundant NICs, multiple subnets, and at least four switches and storage hosts between production and DR sites. Power redundancy includes UPS and generators. Robust operational procedures, permission controls, and regular DR drills are essential, especially for large telecom operators and banks that follow strict regulatory guidelines.

The author notes that the article is a work‑in‑progress and invites discussion to refine the understanding of database HA strategies.