Tiered Tables

Suggest edits

Tiered Tables enable EDB Postgres® to manage large, time-based datasets efficiently by automatically moving "cold" data to cost-effective object storage, while keeping "hot" data in primary transactional storage.

This pattern optimizes both performance and cost while preserving full analytical access to the entire dataset.

For details on how Tiered Tables are implemented and managed within Hybrid Manager (HM), see Implementing Tiered Tables in Hybrid Manager.

What are Tiered Tables

Tiered Tables are a native capability of EDB Postgres Distributed (PGD), supported by the Analytics Accelerator architecture.

They automatically offload older partitions of time-partitioned tables from PGD to object storage (Apache Iceberg format) using:

PGD AutoPartition for automated partitioning and lifecycle control
PGAA and PGFS for querying and accessing offloaded data
Optional Iceberg catalogs for governance and interoperability

The result: seamless, transparent access to data across hot (PGD) and cold (object storage) tiers.

Related concept: Data tiering

Why Tiered Tables matter for EDB analytics

Tiered Tables help organizations:

Keep PGD operational storage lean and performant
Lower storage costs by offloading old data to object storage
Maintain unified query access to the full dataset
Support both OLTP and OLAP use cases on Postgres
Implement lakehouse architectures for historical analysis

Related concept: Analytics Accelerator concepts

How EDB implements Tiered Tables

Core components:

PGD AutoPartition:
Creates new time-based partitions automatically
Defines analytics_offload_period to control offload timing
PGFS:
Provides access to object storage for offloaded data
PGAA:
Enables unified querying across PGD and Iceberg tiers
Creates an offloaded view (table_offloaded) for cold data only
Optional Iceberg catalog:
Supports governance and cross-platform interoperability

Query behavior:

Queries on the parent PGD table automatically access both hot and cold data.
The PGD query planner pushes WHERE clauses to optimize access across storage tiers.

Related concepts:

Common use cases

Use case	Tiered Tables + Analytics Accelerator
IoT and telemetry	Manage large time-series datasets with automated offload
Regulatory and financial data retention	Cost-efficient storage of historical data with full auditability
Analytical reporting on historical data	Use Lakehouse nodes to query offloaded data at scale
Hybrid OLTP / OLAP patterns	Keep current data fast on PGD, analyze large history in Iceberg