Overview of Database High Availability Technologies: HADR, HACMP, Data Replication, Storage DR, and DPF

Database high availability is a complex systems engineering discipline; this article introduces several fundamental HA technologies such as HADR, HACMP, data replication, storage‑layer disaster recovery, and DPF HA, discussing their applicable scenarios and technical characteristics.

1. DB2 HADR – High Availability Disaster Recovery is IBM DB2’s database‑level HA data‑replication mechanism, originally from Informix. It uses a primary‑standby pair, with read‑only standby support introduced in version 9.7. Synchronous transmission is recommended when bandwidth is stable and loss‑less data is required; asynchronous or super‑asynchronous modes are used for long‑distance links.

HADR cannot be used with DPF and is limited to single‑partition databases, but it generally offers faster failover than Oracle Data Guard (DG) with fewer failure risks. It lacks built‑in compression, encryption, and heterogeneous database replication, though third‑party SSH tools can be integrated.

2. SQL Replication and Q Replication – SQL replication is suited for same‑LAN environments, while Q replication works better over remote links by buffering data via WebSphere MQ. Q replication, often combined with HADR, provides efficient, low‑impact transaction‑log‑based remote data copy, primarily for DB2; Oracle support is limited. CDC (formerly Data Mirror) offers multi‑database table‑level replication, and Oracle GoldenGate is a comparable solution.

3. HACMP – HACMP offers several clustering modes: Cascading (primary‑backup with priority), Rotating (equal‑priority nodes), and Concurrent (no primary/backup, all nodes run resources). These modes address different performance and cost requirements, with Rotating favored for high‑availability telecom services and Concurrent often paired with Oracle RAC or OPS.

4. DPF High‑Availability – DPF itself lacks a dedicated HA solution, but multi‑node configurations can provide limited disaster tolerance if critical catalog nodes remain up. Node failures due to OS or network issues can be recovered via HACMP; database corruption on a node cannot be recovered, making DPF unsuitable for critical OLTP without additional HA measures.

5. Storage‑Layer Disaster Recovery – Techniques include disk mirroring, SRDF (Symmetrix Remote Data Facility) with synchronous, near‑synchronous, or asynchronous modes, and Veritas BMR for full‑system backup and recovery. Choices depend on distance, bandwidth, and cost considerations, ranging from direct fiber links for short distances to DWDM solutions for long‑haul replication.

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 involves UPS and generators. Robust operational procedures, access controls, and regular failover drills are essential, especially for large telecom operators and banks.

The author acknowledges the article is a work‑in‑progress and invites discussion to refine the presented concepts.