Enhanced scalability, security, and AI in Azure Database for PostgreSQL

The pace of AI is continuing to accelerate, and proof-of-concepts apps designed to validate AI functionality are now being pushed into production at scale. As usage grows, developers are being challenged to find scalable solutions for storing and processing large amounts of vector data while also sustaining the accuracy of results for effective usage.

Today at Microsoft Ignite, we’re announcing an array of new features for Azure Database for PostgreSQL Flexible Server that enhance the scalability, security, and manageability of running enterprise workloads, as well as new AI features to improve the accuracy of retrieval augmented generation (RAG) based AI applications.

Cluster scale compute and storage comes to Flexible Server

As AI app development expands into more scenarios, Postgres has become the database of choice for developers due to its support for vector indexes and built-in similarity search. These key AI features, combined with a familiar SQL interface and robust relational engine, make it simple for developers to leverage their existing skills to build this new class of applications. In discussions with developers, we’ve learnt that while simple proofs of concept demonstrate the value of applying AI to applications, the transition to production is a challenge due to the scalability of working with vectors.

To enhance scalability, we’re pleased to announce the public preview of elastic clusters in Azure Database for PostgreSQL. With elastic clusters, developers can now scale their data horizontally across a cluster of Postgres servers to leverage the scalability of distributed query execution and a shared-nothing architecture. Elastic clusters support both row and schema based sharding and are ideal for multi-tenant or AI applications – or a combination of both. The public preview of elastic clusters is available in a subset of regions today and will roll out globally in the coming weeks.

Optimizing accuracy in AI apps

For developers taking their AI apps from POC to production, we’re pleased to announce a limited public preview of DiskANN. This new vector index, developed by Microsoft, complements the vector type and query operations provided by pgvector and enables cost efficient scaling for larger workloads. Additionally, we’re pleased to announce semantic ranker which further improves query accuracy for Postgres AI apps, and the Apache AGE extension that can enhance search accuracy using GraphRAG. The semantic ranker and GraphRAG capabilities are provided as solution accelerators on GitHub and are available today.

Application specific workload optimization at scale

When Postgres applications inevitably grow, and workload patterns evolve, it’s critical that the database remains optimized for the most efficient resource usage to maintain optimal performance. In May we announced the public preview of automatic index tuning which provides CREATE and DROP index recommendations for Flexible Server databases in Azure. Since its launch, over 80% of the recommendations have been to DROP indexes. This has enabled our customers to use less compute to maintain redundant indexes as data changes, as well as the disk IOPS and space required to write and store them. We’re pleased to announce that automatic index tuning is now generally available.

However, another very common approach to optimizing Postgres workloads is parameter tuning. Tuning Postgres parameters is hard to get right, and changing workloads can result in the need to re-optimize over time. Coming soon, we’ll release the public preview of automatic parameter tuning in Azure Database for PostgreSQL, which leverages an optimization model to tailor Postgres parameters precisely to your workload. This feature, combined with automatic index tuning, can save hundreds of hours of manual effort by automatically optimizing, and re-optimizing, each of your individual Postgres apps.

Enduring investments in data security, protection and integration

And finally, as Postgres continues its strong rate of adoption throughout the enterprise, we’re pleased to announce new data protection, security and integration features for Flexible Server. On-demand backups were a highly requested feature and are now in public preview. Managed Identities are now supported in the azure_storage extension for integrating Postgres securely with Azure storage, and for customers integrating their Oracle data with Postgres, the oracle_fdw extension is now available in public preview also.

Azure Database for PostgreSQL - Flexible server is the Postgres platform for enterprise and AI

Throughout 2024, the Postgres team at Microsoft has continued to innovate across the entire feature spectrum. From enterprise security and improved high availability to autonomous workload optimization and AI features for building new apps, Azure Database for PostgreSQL is ready to run the next generation of enterprise and AI workloads.

We’re excited for our 2025 roadmap and can’t wait to kick off the next round of announcements in the new year. To learn more about Azure Database for Postgres, you can visit our product page, or try Azure and PostgreSQL for free.