EDB launches disaster recovery for WarehousePG analytics

EDB has launched disaster recovery for WarehousePG, aimed at organisations running very large analytical data environments.

The feature is designed to help companies restore a full WarehousePG cluster in a separate environment after a site-level outage. It gives customers a supported way to protect data at petabyte scale and recover operations across on-premises, cloud, or hybrid infrastructure.

The launch reflects growing pressure on companies that use analytics continuously rather than in periodic reporting cycles. In banking and other regulated sectors, warehouse downtime can directly affect operations because analysts, business teams, and automated systems rely on live data to make decisions.

Recovery processes based on nightly backups and logical restores can become unworkable as data volumes grow. At very large scale, backups can leave hours of data exposed, while full restores can take days and compete with production workloads.

Banking example

EDB pointed to a banking deployment handling hundreds of terabytes of business-critical data through a round-the-clock analytical pipeline. In that environment, the existing backup and recovery process had become too slow for the size of the warehouse.

According to EDB, full backups took nearly two days and restores took a similar amount of time. Backup jobs also ran through most of the day, creating operational strain on the live system.

After adopting Disaster Recovery for WarehousePG, the customer set up hourly restore points on separate recovery infrastructure without changing the application architecture above the warehouse. EDB said restore-point replay averaged 16 minutes, down from almost two days.

It also reported that full backup time fell from 44 hours to about 11 hours. Query lock time dropped by nearly 99%, which EDB attributed to a shift from logical backup to physical backup.

Those results relate to a specific customer deployment and may differ in other environments depending on workload and system configuration. Even so, they offer a concrete example of the challenge facing organisations as analytical estates grow into the hundreds of terabytes or beyond.

How it works

WarehousePG is EDB's Postgres-based massively parallel processing data warehouse for large-scale analytics. The disaster recovery addition extends the system beyond built-in high availability, which protects against failures within a cluster through mechanisms such as segment mirrors and a standby coordinator.

The new function addresses a broader risk: the loss of an entire site. It allows the coordinator and all primary segments to recover together to a single point in time, so the standby environment returns as one consistent warehouse.

To do this, EDB uses open source Barman and PostgreSQL write-ahead log data to coordinate recovery across the distributed cluster. The company positioned this as a way for Postgres teams to apply familiar recovery principles at warehouse scale without adding a separate proprietary software layer.

The announcement also highlights an update to WarehousePG Enterprise Manager. Disaster recovery configuration will be added to the existing management interface, which already covers cluster health, query diagnostics, and resource use.

Control and cost

One of the main selling points is control over where the recovery environment runs and when it is brought online. Organisations can choose the region, jurisdiction, and infrastructure for the standby system, a point likely to matter for companies with data residency and sovereignty requirements.

Customers can also set restore-point frequency and continuously archive write-ahead logs to align recovery with business recovery time and recovery point targets. Another option is to run a smaller or lower-cost standby environment rather than maintain a duplicate production-sized warehouse.

That approach may appeal to businesses that want a disaster recovery posture without duplicating every ingestion job into a second live warehouse. It also gives companies a way to separate resilience planning from day-to-day production spending.

EDB said the same foundation could also support cloud migration, testing under peak-volume conditions, and point-in-time restores for heavy reporting periods such as month-end or quarter-end processing. In those cases, teams can run the extra compute temporarily and release it once the work is finished.

WarehousePG sits within the wider EDB Postgres AI portfolio, which the company markets as a platform for transactional, analytical, and AI workloads built on Postgres. For this launch, however, the immediate focus is narrower: how to recover a large analytical warehouse quickly enough when downtime becomes unacceptable.

"At petabyte scale, a recovery process that takes days is not a recovery plan, it is unhedged business risk. WarehousePG brings the full distributed cluster back to a consistent point in time without forcing customers to trade production performance for protection. That moves disaster recovery from a best-effort exercise to a capability teams can test, measure, and trust," said Max Romanenko, Chief Engineering Officer, EDB.