AlloyDB for PostgreSQL is Google’s managed, PostgreSQL-compatible database for enterprise workloads, with high transaction performance, an integrated analytics engine, and AlloyDB AI for vector search.
What is AlloyDB?
AlloyDB is a PostgreSQL-compatible database built for business-critical applications. Unlike standard PostgreSQL, AlloyDB uses Google’s distributed storage architecture and offers higher performance with full PostgreSQL compatibility.
The architecture separates compute and storage. This enables independent scaling of both components and higher availability through distributed storage. Transactional data is automatically mirrored in a columnar engine for fast analytics without ETL.
AlloyDB is wire-compatible with PostgreSQL 14, 15, 16, 17, and 18. PostgreSQL 17 has been available since September 2025 and PostgreSQL 18 since March 2026. Existing applications, tools, and extensions work without changes. Migration from on-premise PostgreSQL or Cloud SQL uses Database Migration Service with minimal downtime.
Core Features
- High transaction performance through Google’s distributed storage architecture with separated compute and storage
- Integrated columnar engine for analytical queries up to 100x faster than standard PostgreSQL, with no separate data warehouse
- AlloyDB AI with the ScaNN vector index and pgvector for generative AI and semantic search
- Full wire compatibility with PostgreSQL 14 through 18, including extensions like PostGIS and pgvector
- Automatic multi-zone high availability and point-in-time recovery up to 35 days
- C4A Axion (Arm) machines for better price-performance on transactional workloads
AlloyDB AI: Vector Search for Generative AI
AlloyDB AI brings vector search directly into the database, without a separate vector store:
- ScaNN vector index: ScaNN is based on vector search research from Google Research and significantly accelerates search compared to pgvector HNSW. The tree-based index scales to 100 million to over 1 billion vectors.
- pgvector: The familiar pgvector extension remains fully supported, including HNSW indexes.
- In-database embeddings: Embeddings can be generated and queried directly from SQL, which simplifies RAG pipelines and recommendation engines.
This lets you consolidate transactional data and AI workloads in a single database, without replicating data into a separate system.
Typical Use Cases
- Enterprise OLTP: Business-critical applications with high performance and availability requirements.
- Hybrid Transactional/Analytical Processing (HTAP): Real-time analytics on production data via the columnar engine, without a separate data warehouse or ETL.
- Generative AI and semantic search: RAG pipelines, recommendation engines, and vector search with ScaNN and pgvector directly in the database.
- PostgreSQL migration: Move from on-premise PostgreSQL or Cloud SQL without code changes.
- SaaS platforms: Multi-tenant architectures with high scaling requirements.
Benefits
- High performance without lock-in: More throughput than standard PostgreSQL with full wire compatibility.
- One database for OLTP, OLAP, and AI: The columnar engine and AlloyDB AI replace separate analytics and vector systems.
- Managed operations: Automatic backups, point-in-time recovery, multi-zone HA, and a 99.99% SLA including maintenance windows.
- EU data sovereignty: Deploy in EU regions such as Frankfurt, Belgium, and the Netherlands, with encryption and customer-managed encryption keys via Cloud KMS.
- Flexible pricing: Resource-based billing with committed use discounts and better price-performance through C4A Axion machines.
AlloyDB Omni: Hybrid and Multi-Cloud
AlloyDB Omni runs AlloyDB on Kubernetes outside Google Cloud: on-premises, on other clouds (AWS, Azure), at the edge, or in air-gapped environments. It offers identical APIs to Cloud AlloyDB, including the columnar engine and AlloyDB AI vector search. Omni requires your own Kubernetes cluster management and forgoes Google Cloud managed services such as automatic backups, but provides full control and flexibility for hybrid architectures and data residency requirements.
Integration with innFactory
As a Google Cloud partner, innFactory supports you with AlloyDB: migration from on-premise or Cloud SQL with Database Migration Service, performance tuning and columnar engine configuration, HA/DR setup with multi-zone HA and cross-region replicas, AlloyDB AI and vector search design, and hybrid deployments with AlloyDB Omni. Contact us for advice on AlloyDB and PostgreSQL on Google Cloud.
Available Tiers & Options
AlloyDB Primary
- High transaction performance vs standard PostgreSQL
- Columnar engine for analytics without ETL
- AlloyDB AI with ScaNN vector index
- Automatic backups and point-in-time recovery
- C4A Axion (Arm) for better price-performance
- Higher cost than Cloud SQL
- Regional only (no multi-region writes)
AlloyDB Omni
- On-premises, other clouds, or edge
- Kubernetes-based with operator
- Identical API and AlloyDB AI as Cloud AlloyDB
- Self-management required
- No Google Cloud managed services
Typical Use Cases
Technical Specifications
Frequently Asked Questions
What is AlloyDB and how does it differ from Cloud SQL PostgreSQL?
AlloyDB is a fully managed PostgreSQL-compatible database specifically optimized for enterprise workloads. Compared to Cloud SQL PostgreSQL, AlloyDB offers up to 4x higher transaction performance and 100x faster analytics through the integrated columnar engine. AlloyDB uses Google's distributed storage architecture for higher availability and scalability.
What is the AlloyDB Columnar Engine?
The columnar engine is an integrated analytics engine that automatically replicates transactional data in a column-oriented format. This enables OLAP queries (analytics) up to 100x faster than standard PostgreSQL without ETL processes or separate data warehouses. Transactional and analytical workloads run on the same database without performance impact.
Which PostgreSQL versions does AlloyDB support?
AlloyDB is wire-compatible with PostgreSQL 14, 15, 16, 17, and 18. PostgreSQL 17 became generally available in September 2025 and PostgreSQL 18 in March 2026. Existing applications work without code changes. Extensions like PostGIS, pgvector, and standard PostgreSQL tools are supported, which makes migration from on-premise PostgreSQL or Cloud SQL straightforward.
When should I use AlloyDB instead of Cloud SQL PostgreSQL?
Use AlloyDB for business-critical applications with high performance requirements, hybrid OLTP/OLAP workloads, generative AI use cases, or large data volumes over 1 TB. Cloud SQL PostgreSQL is better for smaller workloads, development environments, or when multi-region deployments are needed. AlloyDB offers higher performance, Cloud SQL more flexibility with regions.
Does AlloyDB support vector search for AI applications?
Yes, through AlloyDB AI. In addition to the pgvector extension, AlloyDB offers the ScaNN vector index, based on years of vector search research at Google Research. ScaNN significantly speeds up search compared to pgvector HNSW and scales to 100 million to over 1 billion vectors. This suits RAG (Retrieval-Augmented Generation), recommendation engines, and semantic search directly in the database.
How does AlloyDB pricing work?
AlloyDB uses resource-based pricing: vCPU, RAM, and storage are billed separately, and database storage grows automatically with usage. Committed use discounts with 1-year or 3-year terms reduce cost for predictable workloads. A 30-day free trial cluster with 8 vCPUs is available for testing, and new Google Cloud customers also receive 300 USD in credits. See the official pricing page for current per-region prices.
How does high availability work in AlloyDB?
AlloyDB provides automatic multi-zone high availability within a region, backed by a 99.99% SLA that includes maintenance windows. A primary node handles writes, and Google's distributed storage replicates data synchronously across zones for durability. Cross-region read replicas add geographic redundancy and lower read latency for global applications.
Note: All product information on this page has been compiled with care, but is provided without guarantee and may be outdated or incomplete. Cloud services evolve rapidly — features, pricing, SLAs, and availability change frequently. Authoritative and up-to-date information can only be found on the official product page of Google Cloud (official documentation). This page does not represent an offer by Google Cloud.
