Azure Database for PostgreSQL: AI-Ready for Enterprise Applications in Flexible Server
Published Nov 15 2023 08:00 AM 6,943 Views

Today at Microsoft Ignite, we’re pleased to announce a broad set of new features for Azure Database for PostgreSQL - Flexible Server, Microsoft’s fully-managed Postgres service built on open-source Postgres and engineered for enterprises and developers alike.


The new Flexible Server features enhance the performance, security, and availability of the PostgreSQL database service and also enable you to bring the power of Azure OpenAI directly to the database to simplify the development of new AI-powered apps using the latest version of Postgres.


All of these new features are available in all Flexible Server regions today. In this post, you’ll get an overview of each Azure Database for PostgreSQL announcement at Ignite, plus links to more details via documentation and blog posts.



Postgres 16 support in Flexible Server – General Availability


On September 14th, The PostgreSQL Global Development Group announced the release of PostgreSQL 16. And today, only 2 months after the release of PG16, we’re announcing that PostgreSQL 16 support is now generally available for Flexible Server in Azure Database for PostgreSQL. 



Near Zero Downtime Scaling – General Availability


One of the great benefits of cloud computing is on-demand elasticity. Scaling-up provides an immediate solution to increased usage, while scaling-down enables customers to reduce infrastructure costs during periods of reduced demand. The new near zero downtime scaling feature in Flexible Server in Azure Database for PostgreSQL now ensures that your applications will typically experience less than 30 seconds downtime when resizing compute and/or storage. This enables you to scale more dynamically as your workload changes.



pgvector 0.5.1 support – General Availability


Support for the latest open source version of pgvector, version 0.5.1, is now generally available in Flexible Server in Azure Database for PostgreSQL. The pgvector extension provides the vector data type which can be used for storing AI-generated embeddings. pgvector enables indexing and similarity searches such as cosine similarity. The 0.5.1 version introduces performance improvements for creating HNSW indexes (Hierarchical Navigable Small World) which were introduced in pgvector version 0.5.



New Postgres Extensions: azure_storage & tds_fdw


To simplify loading and exporting data between Azure cloud storage and Postgres, we’re pleased to announce the general availability of the azure_storage extension for Postgres.


Also, to access data in Microsoft SQL, we’re pleased to announce the availability of tds_fdw which enables customers to use a foreign data wrapper to query data stored in Microsoft SQL databases.



New Regions - General Availability


We’re pleased to announce the general availability of the Israel Central region for Azure Database for PostgreSQL - Flexible Server. We are continuing to invest in the global availability of Flexible Server and intend to roll out additional regions throughout 2024.



Premium SSD v2 – Public Preview


Flexible Server in Azure Database for PostgreSQL now provides a new storage option that supports up to 80K IOPS, 64 TB storage, and 1,200 MB/s throughput. The Premium SSD v2 option (in Preview) enables you to tailor IOPS and storage to your workload needs and adjust the settings anytime, without any interruption. You can also resize your storage allocation by as little as 1 GB to achieve the optimal fit for your database.


With the fine-grained control over IOPS and storage that Premium SSD v2 gives you, you can run your most demanding Postgres workloads on Flexible Server on Azure.



azure_ai Extension for Postgres – Public Preview


Big news, Azure Database for PostgreSQL - Flexible Server is now directly integrated with Azure OpenAI. The new azure_ai extension (in Preview) enables you to build highly scalable AI-powered apps entirely within the database layer using the SQL language and Postgres extension model you love. With a few lines of SQL code, you can create embeddings from your data using the power of Azure OpenAI LLM’s, and then store these embeddings in your database. This enables you to implement capabilities such as semantic search, recommendations, and anomaly detection faster and more easily than ever before, and without extensive application changes.


In addition to integration with Azure OpenAI, the new azure_ai extension also enables you to integrate Postgres with Azure AI Language (formerly Azure Cognitive Services) so you can do sentiment analysis, text summarization, language detection, and PII recognition directly in Postgres.



Private Link – Public Preview


Private Link (in Preview) is now available in Azure Database for PostgreSQL - Flexible Server. With Private Link, you can connect to your database using a private IP address within your Azure Virtual Network, without exposing any data to the public internet. This complements the existing secure networking features that require Azure Database for PostgreSQL to be deployed into an Azure Virtual Network.



Multi-Region Disaster Recovery (GeoDR) – Public Preview


Mission-critical applications require robust high availability and disaster recovery capabilities. We’re pleased to announce that Multi-Region Disaster Recovery, often referred to as GeoDR, is now in public preview for Azure Database for PostgreSQL - Flexible Server. Using GeoDR, you can now failover to a different region under a single Virtual Endpoint. How? You can execute a simple “promote to primary” operation from your primary region to a replica in a secondary region, with no connection changes required in the application. Fallback to the original region is equally simple, and the replication is maintained through these operations.



Storage IOPS Separation – Public Preview


Every workload is a little different. Some require very high IOPS on a relatively small database, or alternatively might require small IOPS on a larger database, or anywhere in between. The new IOPS Separation feature (in Preview) for Premium SSD v1 in Flexible Server enables you to optimize your storage configuration to fit your workload, without needing to over-provision storage to meet your IOPS requirements. The result: the same performance at reduced cost.



Long-term Backup Retention – Public Preview


Long term retention is a key component of a comprehensive disaster recovery strategy, and in many cases, is also a regulatory requirement that enables you to meet compliance standards. At Microsoft Ignite 2023, we’re pleased to announce that long-term backup retention for Azure Database for PostgreSQL - Flexible Server is now in public preview. This feature enables you to store Postgres backups of up to 10 years in Azure Backup, and of course restore them at any point in the future. Backups can be configured in the portal or using a programmatic interface such as API or PowerShell.



At Microsoft, we are deeply invested in PostgreSQL


In addition to the fully managed Flexible Server Postgres service in Azure, our Microsoft team also commits and contributes directly to open-source Postgres. This blog post walks you through all of the investments Microsoft has made into Postgres—including open source Postgres over the past 12 months.


It’s been an exceptionally busy year for the Postgres community worldwide and for the Azure Database for PostgreSQL team here at Microsoft. I’m grateful to entire Azure product and engineering team for their commitment and execution over the past year. We have a strong roadmap and we’re looking forward to continuing to deliver great innovations throughout 2024.

Version history
Last update:
‎Nov 15 2023 08:44 AM
Updated by: