AnalyticDB for PostgreSQL is a massively parallel processing (MPP) data warehouse service that delivers real-time analytics on petabytes of data. Compatible with ANSI SQL 2003, PostgreSQL, and Oracle ecosystems, it combines hybrid row-column storage, tiered hot/cold data management, and built-in AI capabilities — including vector search and Retrieval-Augmented Generation (RAG) — in a single engine.
Architecture
AnalyticDB for PostgreSQL is available in two deployment modes:
| Mode | Architecture | Best for |
|---|---|---|
| Elastic storage mode | Shared-nothing architecture based on Elastic Compute Service (ECS) and Enterprise SSDs (ESSDs); provides MPP capabilities | Stable, predictable analytical workloads |
| Serverless mode | Shared-storage architecture based on ECS, local cache, and Object Storage Service (OSS); decoupled storage and compute | Elastic workloads with variable or unpredictable demand |
Both modes share the following two-tier node structure:
Coordinator layer — A coordinator node handles metadata management and load balancing across the cluster.
Compute layer — Multiple compute nodes process data in parallel. Each node runs the Orca optimizer, the Laser execution engine, and the Beam storage engine for high-performance query execution. Incremental materialized views (IMVs) support real-time data warehouse scenarios.
Storage layer — Hot data is stored on ESSDs attached to the compute nodes; cold data is offloaded to OSS. This tiered model improves query performance while reducing storage costs. Storage capacity can be scaled up but cannot be scaled down, ensuring data persistence and stability as your data grows.
Key capabilities
Flexible SQL compatibility. AnalyticDB for PostgreSQL is fully compatible with ANSI SQL 2003 and partially compatible with Oracle syntax, including PL/SQL stored procedures. Next-generation query optimizers eliminate the need to manually tune complex SQL statements.
Petabyte-scale analytics. The MPP scale-out architecture responds to queries across petabytes of data in seconds. Combined with vector computing and intelligent column store indexing, query performance is approximately 10 times higher than a traditional database engine. A single engine supports batch processing, stream computing, and interactive analysis.
High availability. Distributed transactions, full node and data redundancy, automatic failover, and atomicity, consistency, isolation, durability (ACID) compliance ensure always-on connectivity.
Broad ecosystem integration. Built-in support for mainstream business intelligence (BI) and extract, transform, load (ETL) tools. The PostGIS extension enables geographic data analysis; the MADlib library provides over 300 built-in machine learning algorithms.
Flexible data ingestion. Ingest data in real time or in batch from a wide range of sources using Data Transmission Service (DTS) and DataWorks. High-concurrency access to OSS enables direct data lake analysis without data movement.
AI and vector capabilities. Run vector search, build one-stop RAG pipelines, create enterprise knowledge bases, and perform text-to-image and image-to-image search — all within the same data warehouse, without moving data to a separate platform.
Limitations
For a full list of service limits, see Limits.
Next steps
| If you are... | Start here |
|---|---|
| Evaluating whether AnalyticDB for PostgreSQL fits your use case | Review Key capabilities and Architecture, then check Limits |
| Ready to create your first instance | See the Quick Start guide |
| Looking to migrate an existing data warehouse | See the Data migration guide |
| A developer integrating via SQL or SDK | See the Developer guide |
Technical support
To get help from the community, join the AnalyticDB for PostgreSQL DingTalk group 11700737.