AzureFeeds

In this post you’ll find an overview of all the Postgres work happening at Microsoft—both in the managed Azure database service and in the upstream open source project. I’ve pulled together highlights from each workstream to show what’s new over the past 12 months.

There’s been a lot of progress. You probably already know that Postgres is the world’s most advanced open source relational database, trusted by developers and used everywhere from Fortune 500 enterprises, to small and medium-sized businesses, to individual developers working on side projects. But here’s something you might not have heard: just a few days ago, Postgres came up four times in Microsoft’s third-quarter earnings call. And Postgres on Azure is now powering technologies like ChatGPT, at global scale.

The table of contents below follows the structure of this hand-made infographic. Compared to last year’s blog post, I had to make this infographic taller—there’s just that much more to cover. You can jump straight to the sections you care about, or settle in ☕️ and read the whole thing. Enjoy!  

• Azure Database for PostgreSQL – Flexible Server


• PostgreSQL core


• Citus open source


• Postgres community work

 

Figure 1:This hand-made infographic maps out the major workstreams across Microsoft’s Postgres efforts—from the managed service on Azure, to upstream contributions to the Postgres core, to the Citus extension, and contributions to the Postgres community. It’s a snapshot of just a subset of the work we’ve delivered over the last 12 months. For much of the open source work depicted, our contributions to the PostgreSQL open source project involved collaboration with people across the global PostgreSQL community, both inside and outside Microsoft.

Azure Database for PostgreSQL – Flexible Server

Azure Database for PostgreSQL – Flexible Server is the name of our managed database service on Azure.

Flexible Server meets the needs of large enterprises, small to medium-sized businesses, and early-stage startups—and has been growing in capability as evident in the monthly “feature recap” blog posts on Microsoft Tech Community over the last year (props to Varun Dhawan for publishing these) in May 2024; June; July; August; September; the November Ignite recap from Charles Feddersen; January 2025; and March 2025.

HEADS-UP: Throughout this post, I sometimes use the full name—Azure Database for PostgreSQL – Flexible Server—and sometimes I just say “Flexible Server” for short. They mean the same thing. I noticed a bit of back-and-forth between the two names while proofreading, so am flagging it here to avoid any confusion.

New AI capabilities in Flexible Server

• Azure AI Agent Service support (Preview): The Azure AI Agent Service support (part of Azure AI Foundry) enables you to build AI agents that interact with Azure Database for PostgreSQL—helping you combine LLMs with Postgres data to perform tasks like querying, analyzing, and summarization. (link to blog post on how to build AI agents)
  
  


• DiskANN vector index (Preview): DiskANN is a vector search extension that uses an approximate nearest neighbor search algorithm—and is available via the pg_diskann extension in Azure Database for PostgreSQL. Created by Microsoft Research, DiskANN currently surpasses the HNSW and IVFFlat vector indexes in terms of latency and accuracy. (link to Abe’s DiskANN blog post)
  
  


• pgvector 0.8.0 support: When new versions of the open source pgvector extension for Postgres (also referred to as vector) become available, our team adds support for it into Flexible Server. As of the writing of this blog post the latest version, pgvector 0.8.0, is supported in Flexible Server. (link to docs on how to enable and use)
  
  


• azure_ai extension: The azure_ai extension to Azure Database for PostgreSQL became generally available (GA) in May 2024. With azure_ai, you can integrate Azure AI services—including Azure OpenAI and Azure Machine Learning —directly within Postgres using SQL queries, simplifying your generative AI application development and supporting things like real-time scoring with ML models, sentiment analysis, summarization, and other LLM services. (link to docs)
  
  


• Apache AGE extension support (Preview): AGE is an acronym for “A Graph Extension”. The AGE open-source extension support in Azure Database for PostgreSQL enables you to implement the GraphRAG pattern, so you can store and query graph data efficiently within Postgres. It supports the openCypher query language. (link to Josh’s blog post announcement)
  
  


• Model Context Protocol (MCP) Server (Preview): The new MCP support in Flexible Server lets your AI application interact with your database using a standardized protocol, so your application’s users can run natural language queries to interact with the data in Postgres. (link to Shreya’s blog post announcement)
  
  


• Semantic Ranker solution accelerator: This solution accelerator extends Azure Database for PostgreSQL so you can perform semantic ranking directly via SQL queries. It uses a SQL function and an Azure Machine Learning model—plus the azure_ai extension—to compare text and boost the accuracy of search results, making it easier to find relevant data in apps like AI-driven search or recommendation systems. (link to Semantic Ranker solution accelerator)
  
  


• GraphRAG solution accelerator (Preview): GraphRAG is an advanced retrieval-augmented generation (RAG) technique for storing and querying graph data. GraphRAG was created by Microsoft Research, made possible in Postgres thanks to the Apache AGE extension. This new solution accelerator gives you an end-to-end example of applying the GraphRAG technique to a legal research application to boost the quality and accuracy of LLM responses. (link to GraphRAG solution accelerator)

Migration service in Azure Database for PostgreSQL

• More database migration sources: The Migration service is designed to assist you in moving your Postgres databases to Flexible Server, giving you support for schema and data migrations at unlimited database size. And now it supports offline and online migrations from more sources, including Amazon Aurora for PostgreSQL and Google Cloud SQL for PostgreSQL. (link to docs for Migration service)
  
  


• Migration support for extensions: In response to popular requests, the Migration service now supports additional PostgreSQL extensions—specifically: pg_partman, postgis_topology, timescaleDB, and postgis_tiger_geocoder. The Migration service support includes migrating configuration tables defined by these extensions, preserving critical metadata and settings during migration, similar to pg_dump. (link to docs)

Performance & Scalability

• Reserved pricing for Intel & AMD V5 SKUs: By reserving compute resources in advance, you get significant savings over pay-as-you-go pricing options. Commit to either a one-year or three-year plan to receive up to 65% off on compute costs. (link to blog post)
  
  


• Elastic clusters (Preview): Elastic clusters is a newly-introduced feature integrated into Flexible Server that uses the Citus open source extension to give you horizontal sharding (both row-based and schema-based). Many customers have asked that Citus be integrated into Flexible Server and finally here you have it. (link to Adam’s announcement blog post / link to docs)
  
  


• Fabric mirroring (Private Preview): With Fabric mirroring in Azure Database for PostgreSQL, you get seamless integration of your operational data into Microsoft Fabric, so you can take advantage of Fabric’s advanced analytics. Fabric mirroring supports near real-time replication of your PostgreSQL data, which is incrementally replicated into Fabric OneLake in Parquet format. (link to Silvano’s announcement blog post)
  
  


• Automated index tuning: This built-in feature in Flexible Server automatically analyzes tracked queries and recommends when indexes should be created, dropped, or rebuilt. Also included are several settings that give you controls on the behavior of the optimization algorithm. (link to Nacho’s announcement blog post)
  
  


• IOPS scaling: Gives you the ability to dynamically adjust your IOPS to meet the demands of your workload, so you can fine-tune performance and manage your costs, without having to overprovision resources. (link to docs)
  
  


• Storage autogrow for read replicas: The storage autogrow feature allows database storage to increase automatically to prevent running out of space—and as of July 2024, read replicas are now supported with storage autogrow. (link to docs)

Security

• Versionless CMK (Preview): Versionless customer managed keys (CMK) make your life better by simplifying your key rotation. If you rotate keys in Key Vault (regardless of whether auto-rotation or manual rotation), you will appreciate this new capability, since the server will automatically use the new version of the key for encrypting and decrypting data. (link to docs)
  
  


• ADF Connector Entra support: A new Azure Data Factory (ADF) connector with enhanced security features was added in March 2025, supporting TLS 1.3, multiple SSL modes, and Service Principal (Entra) authentication. This connector works when building data pipelines to Flexible Server from either ADF or from Azure Synapse. (link to docs)
  
  


• CAST function improvements: The CAST function is used to convert data types within SQL queries and in 2024 we addressed previous limitations with non-binary-coercible casts, making it easier to manage data integrity during data transformation and during migrations to the cloud. (link to blog post announcement)
  
  


• Managed Identity in azure_storage: The azure_storage extension allows you to import or export data in multiple file formats directly between Azure Storage accounts and Azure Database for PostgreSQL – Flexible Server. And now Managed Identity is supported in azure_storage, too. (link to docs)
  
  


• System Assigned Managed Identity: This feature allows you to use Flexible Server’s identity to interact with other Azure services—and was a prerequisite for integrating Azure Database for PostgreSQL into the Fabric ecosystem. (link to docs)

QuickStarts

• Java & .NET SDK QuickStarts: With these new QuickStart guides for the Azure SDK for Java—and the Azure SDK for .NET—you can learn how to create, update, and delete an Azure Database for PostgreSQL flexible server instance. Useful for both experts and new users. (link to java sdk quickstart / link to dotnet sdk quickstart)

Backup

• On-demand backups: In addition to your scheduled automatic backups of Azure Database for PostgreSQL, you can also create an on-demand backup whenever you choose—and which can be retained according to your backup retention window. (link to docs / link to Kanchan’s blog post announcement)
  
  


• RPO reduced to 5 minutes: The Recovery Point Objective (RPO) for Flexible Server was updated to reduce it from 15 minutes to just 5 minutes on systems without high availability configured. (link to business continuity docs)
  
  


• CMK support for backup LTR: Azure Database for PostgreSQL – Flexible Server and Azure Backup have built an enterprise-class backup solution with long-term retention (LTR) that retains backups for up to 10 years. And in June 2024 we rolled out CMK support for long-term backup retention (LTR). (link to docs)

Manageability

• Database size metrics: The database size metric in Flexible Server gives you precise, granular insight into the storage consumption of individual databases—helping you with capacity planning. (link to blog post announcement)
  
  


• HA health status monitoring: Leveraging the Azure Resource Health Check (RHC) framework, this service gives you continuous insights into the health of your HA-enabled instances and notifies you of events that might affect connectivity and/or availability. (link to docs)
  
  


• pg_signal_autovacuum_worker: Available in Postgres versions 15 and higher, the pg_signal_autovacuum_worker role lets non-super users terminate autovacuum. (link to Gauri’s excellent blog post explainer)
  
  


• New Azure Advisor recommendations: Azure Advisor is a cloud assistant that analyzes your Flexible Server configuration and usage telemetry to give you personalized recommendations. New Advisor recommendations include checks for prepared transactions that are long-running &/or orphaned. (link to docs)
  
  


• Automation tasks: Automation tasks are more basic and lightweight vs. Azure Automation. An automation task is a logic app resource that runs a workflow powered by the Azure Logic Apps Service—and you can use automation tasks to start or stop Flexible Servers on a predefined schedule. (link to docs)
  
  


• Built-in Azure Policies: Azure Policy helps you enforce organizational standards and assess compliance at scale, giving you an aggregated view via a compliance dashboard, along with granular drill downs. (link to docs)
  
  


• Server parameter enhancements: Per Varun’s blog post announcement, we enhanced several server parameters to boost database performance and customizability, notably the ability to adjust settings for vacuum behavior, memory allocation, and client connection defaults. (link to docs)
  
  


• 5 new regions added: Over the last 12 months we’ve added support for 5 more regions for Flexible Server—specifically New Zealand North, Mexico Central, Spain Central, China North 2, and China East 2—bringing the total at the time of this post’s publication up to 58 regions. The latest region support can always be found in the Azure docs. (link to docs)

Versions & Extensions

• Postgres 17 support (Preview): Postgres 17 support was added to Azure Database for PostgreSQL in preview within just a week of the PostgreSQL 17 GA, and in the future it will transition to GA support along with major version upgrade support. (link to Varun’s blog post announcement)
  
  


• Postgres 16 major version upgrade: Support for in-place major version upgrade to Postgres 16 was made generally available back in June 2024, supporting both the Azure portal and the Azure CLI. (link to docs)
  
  


• Latest PostgreSQL minor version support: The PostgreSQL global development group usually releases new minor versions each quarter—and once we’ve QA’d the new minor releases, Flexible Server automatically performs minor version upgrades to the Azure-preferred PostgreSQL version as part of periodic maintenance. As of the date of publication of this post, Flexible Server supports Postgres 17.4, 16.8, 15.12, 14.17, and 13.20. (link to release notes)
  
  


• 8 extensions added: Support for 8 new Postgres extensions was added to Flexible Server over the last 12 months, including postgres_protobuf, login_hook, session_variable, hll, topn, plus the oracle_fdw, age, and pg_diskann extensions. (link to extensions docs)
  
  


• 24 extensions updated: The support in Flexible Server for 24 extensions was updated in this timeframe to add support for newer versions, including timescaledb, anon (preview), pgvector, and pg_diskann, among others. (link to supported extensions)
  
  


• ALTER extension UPDATE: Support for this command makes it easy for you to manually upgrade to the latest version of an extension that is supported by Azure. (link to docs)
  
  


• oracle_fdw extension support: The oracle_fdw extension is a foreign data wrapper for Oracle databases, giving you the ability to interact with Oracle data sources from within your PostgreSQL databases. (link to Alper’s announcement blog post)

Try it out: Free trial for Azure Database for PostgreSQL

Same guidance as previous years: the “Use an Azure free account” docs page walks you through how to get started with Burstable B1MS instance—that’s 750 hours per month, with 32 GB of storage and 32 GB of backup storage, free for the first 12 months.

Our team is hiring!

If you’re a deep Postgres expert—or a strong Linux systems engineer—this might be your moment. As Affan Dar, our VP of Engineering, shared in a recent LinkedIn post, we’re growing the Postgres team at Microsoft.

“We are always looking for strong Linux systems engineers and deep PostgreSQL experts to help us build the future of Postgres at Microsoft.”
—Affan Dar, VP Engineering at Microsoft

Contributing to PostgreSQL core

Microsoft has been building a team of Postgres open source contributors since 2019.

Postgres is widely used—by millions of people around the world, in the cloud and on-prem. But the future of Postgres needs ongoing investment to stay technically and commercially relevant. That’s why we’re continuing to invest in our open source contributor team, as KK shared in his recent blog post reflecting on 2024.

The TL;DR? Our commitment to supporting Postgres development hasn’t changed since I first wrote this back in August 2023:

In order to thrive, an open source ecosystem needs commercial support as well as volunteer efforts. Even open source developers need to eat! For the Postgres open source ecosystem to flourish, companies like Microsoft need to support the project by funding development in the Postgres core. Which we do.

PostgreSQL is a complex piece of software that runs mission-critical workloads across the globe. To provide the best possible experience on Azure, it follows that we need to thoroughly understand how it works. By having PostgreSQL committers and contributors on our team, they can share knowledge internally across different orgs, or directly answer internal questions regarding incidents or extension development.

Because today’s cloud operates at a scale most on-prem solutions never encountered, unique cloud data center problems, often relating to performance, now require special attention. Our in-house team of deep Postgres experts are focused on tackling these cloud-scale issues upstream, in the Postgres core. Another benefit: our team’s Postgres expertise gives Azure customers confidence in our cloud database services, too.

Commercial funding of PostgreSQL developers has another benefit: it gives developers the long-term stability to pursue the big things, the groundbreaking changes that are super important to the future. In particular, the Postgres contributor team at Microsoft is focused on some big architectural changes (example: Asynchronous IO) that you wouldn’t be able to do without the funding for a full-time, multi-year effort.
—Claire Giordano, Head of Postgres open source community initiatives at Microsoft

Highlights of our team’s Postgres 18 code contributions (so far!)

There is a new Postgres major release each year—usually in September or October—and the feature freeze is usually in April. So while Postgres 18 is not yet finished, we have sufficient insight into what will likely be in Postgres 18 to share some highlights of the Microsoft team’s code contributions.  

First, some high-level metrics about Microsoft’s contributions to the Postgres 18 core release (to date, as of 28 April 2025) are:

 

To interpret the numbers, here are 3 things to keep in mind:

• Not all commits are created equal. Some commits are small; others are complex and high-impact. So while commits are a useful signal, they’re not a perfect measure of effort or impact. Still, it’s a metric many people care about, so we’re sharing it.
  
  


• Many commits have multiple authors (and reviewers): The 456 commits with Microsoft authors in Postgres 18 (as of Apr 28, 2025) were authored by 13 different people on our team. But those commits were often co-authored by other contributors—both inside and outside Microsoft. That’s one of the things I love about the Postgres community: developers collaborating across companies, countries, and continents.
  
  


• Code is just one part of the story: Code and review contributions are essential to the Postgres project—but they’re not the whole picture. As you’ll see later in this post, our team also contributes in other ways: to extensions like Citus, and to the Postgres community.

Figure 2: Photos of the PostgreSQL open source hackers at Microsoft who focus primarily on the Postgres core. In addition to this team, other developers on our broader Postgres team also contributed to Postgres 18 core—including Ayush Tiwari, Nitin Jadhav, and Vinod Sridharan. We also have contributors working on the broader ecosystem (like the Citus team and Alexander Kukushkin’s work on Patroni), as well as others—myself included—who support the Postgres project in non-code ways.

Now let’s take a look at some highlights from the 450+ commits our Microsoft team authored or co-authored (so far) in Postgres 18.

Attribution matters in open source. So it’s worth saying clearly: the work you’ll see below was done in collaboration with Postgres developers and contributors from around the world—including folks outside Microsoft. That’s how open source works. And we’re grateful to be part of it.

Asynchronous IO

• Asynchronous I/O subsystem: Postgres 18 adds significant new infrastructure to perform efficient filesystem I/O with modern operating system features. New to Postgres 18, all read streams will now use true asynchronous I/O with Linux-native io_uring or portable worker processes. In future releases of Postgres, other types of I/O will also benefit from the new async I/O subsystem—which is to say, writes will be supported in the future, too.
  
  When talking with Postgres committer Thomas Munro to learn more about the Asynchronous I/O project for the blog post, Thomas had this to say:
  
  
  

  
  

  “Seeing the Async I/O subsystem land in Postgres 18 is really cool. It’s a big step forward—a foundation for a whole new I/O architecture that opens the door to modern and efficient I/O technologies. It’s been awesome to be able to help Andres Freund with this project (I worked on portability, vectored I/O, streaming APIs, and a few other bits and pieces) as finding a path to land these changes incrementally across releases has been a challenge. And we’re not done yet. The interesting work continues.”
  —Thomas Munro, PostgreSQL hacker @ Microsoft

  
  

  
  
  Links to commits:
  
  
  
  • link to core AIO infrastructure commit
  
  
  • link to higher level design README commit
  
  
  • link to pg_aios commit
  
  
  • link to pgaio batchmode commit
  
  
  • link to read_stream commit
  
  
  • link to StartReadBuffers commit
  
  
  • link to bufmgr AIO read commit
  
  
  • link to smgr md fd commit for AIO
  
  
  • link to io_method io_uring commit
  
  
  • link to liburing dependency commit
  
  
  • link to io_method worker commit for AIO
  
  
  • link to infrastructure for IO workers commit
    
    
  
  
  
  


• Streaming reads in vacuum & bitmap heap scan: Continuing the effort to streamify parts of Postgres that can benefit from I/O combining and asynchronous I/O, Postgres streamifies VACUUM and Bitmap Heap Scan. Links to commits:
  
  
  
  
  • link to VACUUM first phase
  
  
  • link to VACUUM third phase
  
  
  • link to BitmapHeapScan commit
  
  
  • link to AIO batchmode commit
  
  
  • link to scan bitmap next block commit
  
  
  
  

Security & Access

• OAuth infrastructure for authentication: Support for OAuth 2.0 authentication, allowing users to log into the database using secure, token-based credentials from identity providers. The big deal is this: this OAuth work extends the existing framework to handle asynchronous authentication in Postgres, which is the groundwork for future authentication implementations with corporate or third-party accounts. (link to commit)

Scalability Futures

• Early groundwork for Multithreading: Ongoing work to remove obstacles to a future multithreaded Postgres instead of the current multi-process architecture (link to commit / link to commit / link to commit)

System Performance

• Amortize aggressive vacuum cost: Aggressive vacuums often incur large amounts of extra I/O at the expense of other workloads. To amortize the cost of aggressive vacuums—so these aggressive vacuums do not take up so many system resources—this commit enables Postgres 18 to eagerly scan all-visible pages during normal vacuums. (link to commit)
  
  


• Fast database cloning: This is really cool. Postgres 18 introduces a new file_copy_method setting which can be set to either COPY (the default) or to CLONE if the operating system supports it. Some filesystems offer a fast way to copy files using copy-on-write or I/O pushdown. In Postgres 18, the new file_copy_method option allows the copy_file_range system call to be used, which can clone a database nearly instantly on some filesystems such as ZFS, XFS, BTRFS, and APFS. (link to commit)

Query Performance & Optimization

• Reduce planning time for partitioned tables with many partitions: These optimizations are a game changer for applications with large numbers of partitioned tables. If you’re a user and are wondering “will I benefit?”, the answer is: “it depends.” Initial benchmarks with 56 partitions show the planner to be 3X faster. In use cases with thousands of partitions, the benefits are shown to be even greater. (link to EquivalenceMember commit / link to derived clause commit)
  
  


• Optimize memory management for materialize, window aggregation, & CTE scans: Part of a collection of performance optimizations in Postgres 18, this commit optimizes WindowAgg’s use of tuplestores which can yield some dramatic performance improvements when using PARTITION BY. This improvement is especially useful when each window partition only has a small number of tuples. (link to commit)
  
  


• Speed up hash GROUP BY & Hash Joins: Add expression evaluation support for hashing, allowing just-in-time (JIT) compilation to obtain hash values. Initial testing has shown up to ~10% query performance increases with a large number of lookups on a small hash table. And hashed GROUP BY queries have seen performance increases of ~15%. (link to hash joins commit / link to GROUP BY commit)
  
  


• Remove redundant GROUP BY columns using UNIQUE indexes: In Postgres 18, the executor is now more efficient in certain scenarios, saving hashing effort when there are functionally-dependent GROUP BY columns and giving the planner more flexibility to use pre-sorted results from btree indexes. (link to redundant GROUP BY commit)
  
  


• Refactor executor to allow inlining of core logic: Based on initial benchmarks (YMMV), this refactoring leads to a 5% reduction in execution time for queries that rely heavily on sequential scans. (link to commit)
  
  


• Fast path locking improvements: Increase the number of fast-path locking slots. This reduces contention between Postgres sessions for queries that obtain a large number of locks. The previous limitation of 16 fast path slots was always rather low. This change is useful for queries to partitioned tables &/or queries with large numbers of JOINs. (link to commit)

Monitoring & Diagnostics

• Improved Memory observability: Introduces a new pg_get_process_memory_contexts function to get memory context statistics for processes. This new feature will help Postgres users who are struggling to debug situations with memory pressure and want to figure out what is causing the high memory consumption. (link to commit)
  
  


• log_connections improvements: The first improvement modularizes the log_connections output to give users more control over the volume and kind of logging. In addition, a new log_connections option setup_durations has been added to log durations of several key parts of connection establishment and backend setup, to provide visibility into authentication and fork duration as well as end-to-end connection establishment. (link to modularize output commit / link to duration logging commit)
  
  


• Enhanced pg_stat_checkpointer stats: First introduced in Postgres 17, pg_stat_checkpointer is useful for observing and tuning performance of checkpoints. Postgres 18 fixes inconsistent reporting of checkpointer stats. (link to commit)

Developer & Testing Tools

• CI/CD infrastructure maintenance for better test coverage: Ongoing maintenance improvements to the CI/CD tooling used by the Postgres project (Cirrus CI) to test new patches across multiple operating systems.
  
  
  
  • link to FreeBSD commit
  
  
  • link to NetBSD and OpenBSD commit
  
  
  • link to core file commit
  
  
  • link to cached MacPorts commit
  
  
  • link to Debian Bookworm commit
  
  
  
  

Citus Open Source

The Citus open source extension to Postgres is a good fit for data-intensive applications that need more compute, memory, or scale than a single Postgres node can handle.

You can think of Citus as giving Postgres the superpower of distributed tables. The project has over 11K stars on GitHub—and as of November 2024, Citus is now available in Azure Database for PostgreSQL – Flexible Server as a new (preview) feature called elastic clusters.

New Citus open-source features in last 12 months

You can find the release notes for Citus open source releases online—and the latest, Citus 13.0, adds support for Postgres 17. To dig deeper into what’s new in Citus 13.0, you’ve got options:

• Read Naisila’s release blog post titled, Distribute PostgreSQL 17 with Citus 13


• Check out the Citus 13.0 release notes


• Or watch the video of the Citus 13.0 release party livestream

Here are highlights of Postgres 17.2 support added to Citus 13.0 (link to PG17 release notes):

• Optimizer enhancements: One of the Postgres 17 optimizer improvements is relevant to Citus users in particular: the “Allow correlated IN subqueries to be transformed into joins” commit enables Citus 13.0 to plan and execute a query with a correlated IN subquery using query pushdown. (link to the right section of Citus 13 release notes)
  
  


• JSON_TABLE() support for distributed queries: Postgres 17 adds JSON_TABLE() functionality, which triggered the work in Citus 13.0 to perform a join between the JSON_TABLE() function and a Citus distributed table. (link to Citus 13 release notes)
  
  


• Propagate MERGE … WHEN NOT MATCHED BY SOURCE: In the 13.0 release, Citus extended its strategies for handling MERGE in a distributed environment to include the WHEN NOT MATCHED BY SOURCE syntax as well. (link to Citus 13 release notes)
  
  


• EXPLAIN options for MEMORY & SERIALIZE: In Postgres 17, Citus 13.0 now supports the two new options introduced for EXPLAIN. Per the Citus 13.0 release notes: “the SERIALIZE option examines the actual cost of transforming the query’s output into a format suitable for display and the cost of sending that data to the client. And the MEMORY option provides details on the memory usage by the planner. (link to Citus 13 release notes)
  
  


• Distributed partitioned table improvements: Postgres 17 added options for managing data in partitions and Citus 13.0 adds support for propagating these partitioned table options across a distributed Citus cluster. Specifically, options for access method behavior, exclusion constraints, and identity columns. (link to release notes)
  
  


• COPY FROM null options: Citus 13.0 added support for the FORCE_NULL and FORCE_NOT_NULL options in a distributed COPY query. (examples and details in release notes)
  
  


• AT LOCAL operator: Now in Citus 13.0 you can use the AT LOCAL operator in Citus distributed INSERT queries. (link to release notes)

Postgres community work

Given my role leading Postgres open source community initiatives at Microsoft, it’s probably no surprise: contributing to Postgres in ways beyond code is something I care deeply about.

Last year, I gave a talk at PGConf EU that shared an analysis of contributions to Postgres v17, based on research I did with Daniel Gustafsson. And if you’ve ever raced sailboats or been part of a competitive sports team, you know: every role matters. 

To help shine a light on the many kinds of work that keep the PostgreSQL project going, I created the chart below. It was inspired by a pie chart from Josh Berkus—core team emeritus for PostgreSQL—that was part of Josh’s 2007 talk called “50 Ways To Love Your Project”. The goal is to make visible some of the important—but often unsung—contributions that power the Postgres ecosystem.

Figure 3: This pie chart shows the many types of work that go into the PostgreSQL open source project. Code is at the core—but each of these workstreams plays a vital role in delivering new Postgres releases.

Here are a few highlights from the community work our team at Microsoft has contributed over the past year:

• Serve on organizing & talk selection teams for multiple Postgres events: The in-person Postgres community conferences—held all over the world—are a great way to learn, share your expertise, and connect with other Postgres people. And if you lean into the “hallway track”, you might be surprised by the doors it can open.
  
  Over the past 12 months, members of our Postgres team at Microsoft have helped organize or select talks for these 6 conferences:
  
  
  
  
  
  • PGConf.dev 2024 (Melanie Plageman & Daniel Gustafsson)
  
  
  • PGConf NYC 2024 (Daniel Gustafsson)
  
  
  • PGConf.EU 2024 (Tomas Vondra)
  
  
  • Prague Postgres Developer Day (P2D2) 2025 (Tomas Vondra)
  
  
  • PGConf India 2025 (Nikhil Sontakke, Melanie Plageman)
  
  
  • Nordic PGDay 2025 (Daniel Gustafsson)
  
  
  
  


• Co-organize PGConf.dev annual PostgreSQL development conference: Special shout-out to Melanie Plageman and Daniel Gustafsson, who were part of the team that helped revamp the Postgres development conference in 2024—now called PGconf.dev. This annual gathering (the successor to PGCon) is unique in how it brings together current and future PostgreSQL hackers from around the world to collaborate in person. I was also thrilled that our team got to contribute the new PGConf.dev logo design.
  
  Heads-up, PGConf.dev 2025 is happening on 13-16 May in Montréal, Canada—and it’s going to be awesome. This year, Melanie Plageman is once again co-organizing the event. A few registration spots are still available, so if you’ll be nearby, check out the PGConf.dev 2025 schedule—with expert talks from 52 speakers, including 10 from Microsoft—and register while you still can. (link to PGConfdev 2025 schedule)
  
  


• Sponsor 11 Postgres conferences around the world: Postgres community conferences don’t just happen—they rely on financial support to bring people together. Over the past 12 months, our team at Microsoft has sponsored these 11 in-person Postgres events:
  
  
  
  
  
  • PostgreSQL Conference Germany 2024 – Platinum
  
  
  • PGConf.dev 2024 – Gold
  
  
  • PGDay UK 2024 – Benefactor
  
  
  • PGDay Lowlands 2024 – Supporter
  
  
  • PGConf NYC 2024 – Platinum
  
  
  • PGConf EU 2024 in Athens – Platinum
  
  
  • Prague Postgres Developer Day (P2D2) 2025 – Bronze
  
  
  • PGConf India 2025 – Diamond
  
  
  • Nordic PGDay 2025 – Supporter
  
  
  • pgDay Paris 2025 – Partner
  
  
  • PGDay Chicago 2025 – Gold
  
  
  
  Figure 4: Logos of the PostgreSQL community conferences our Microsoft team is happy to have sponsored over the last 12 months—along with thanks and gratitude to the conference organizers for making these events happen.
  
  


• 
  

  POSETTE: An Event for Postgres: POSETTE is a free & virtual developer conference organized by our Postgres team at Microsoft. Now in its 4^th year, it’s the largest Postgres event in the world—bringing together developers, users, contributors, and experts from across the globe to share their knowledge.
  
  Last year’s event (June 2024) was a hit—and you can still watch all the talks on the POSETTE 2024 YouTube playlist. Huge thanks to my teammates on the 2024 talk selection team—Alicja Kucharczyk, Daniel Gustafsson, and Melanie Plageman—for reviewing all the amazing proposals. And of course none of it would’ve happened without Teresa Giacomini, our conference chair, and the full organizing team behind POSETTE 2024.
  
  Looking ahead, POSETTE 2025 is coming up soon: Jun 10-12, online. With 4 livestreams, 2 keynotes, 42 talks, and 45 speakers, there’s something for everyone.

  
  

  Check out the schedule on the PosetteConf website. And add specific POSETTE 2025 livestream(s) to your calendar:

  
  
  
  
  • Livestream 1 / Livestream 2 / Livestream 3 / Livestream 4
  
  
  
  

  Can’t make it live? No problem, you can always catch the talks later on YouTube. But if you attend live, you’ll be able to ask questions in the text chat while the talks are being livestreamed. It’s a good way to connect with the speakers and the community in real time.

  
  
  Figure 5: POSETTE: An Event for Postgres 2025 is a free and virtual developer event, organized by our Postgres team at Microsoft and now in its 4^th year. Formerly called Citus Con: An Event for Postgres, this graphic shows you what the POSETTE acronym stands for: Postgres Open Source Ecosystem Talks Training & Education.
  
  


• Talking Postgres podcast: Hard to believe, but our monthly podcast is now over 2 years old—with 26 episodes published as of April 2025. Focused on the human side of Postgres, the conversations dig into the stories, insights, and experiences of people who shape the project.
  
  In the past year alone, I’ve had the chance to talk with these amazing people:
  
  Bruce Momjian, Dawn Wages, Robert Haas, Daniel Gustafsson, Affan Dar, Andrew Atkinson, Tom Lane, Melanie Plageman, David Rowley, Pino de Candia, Teresa Giacomini & Aaron Wislang, and Michael Christofides.
  
  You can find all episodes at Talking Postgres.com, on the YouTube playlist, or wherever you get your podcasts. Bonus: we record episodes live on the Microsoft Open Source Discord, where you can join the parallel live text chat during the show. It’s a fun way to connect with the community in real time.
  
  


• Deliver 59 conference talks at Postgres events: Both in-person and virtually, our Postgres teams have been sharing their expertise with the user and developer community on the conference circuit—delivering 59 talks in the last 12 months.
  
  


• Organize Postgres user groups & meetups: User groups and meetups are a great way to learn and meet other Postgres users and contributors—but it takes work to make it all happen: to find locations, line up speakers, organize food and drink, and get the word out.
  
  +1 to those on our team at Microsoft—and to everyone—who volunteer their time and partner with others in the Postgres community to help organize local Postgres user groups, including:
  
  
  
  
  
  • Barcelona PostgreSQL User Group: Nacho Alonso Portillo
  
  
  • Czech and Slovak PostgreSQL Users Group (CSPUG): Tomas Vondra   
  
  
  • Madrid PostgreSQL User Group: Nacho Alonso Portillo
  
  
  • Malmö PostgreSQL User Group: Daniel Gustafsson
  
  
  • Stockholm PostgreSQL User Group: Daniel Gustafsson
  
  
  
  


• Serve on Postgres core team, project committees, & PGCA Board of Directors: Andres Freund from Microsoft serves on the PostgreSQL project’s Core Team (and is a prolific committer and major contributor to PostgreSQL as well). Melanie Plageman serves on the PostgreSQL Contributors committee. And as of May 2024, I now serve on the board of directors for the PostgreSQL Community Association (PGCA), a non-profit organization that stewards the PostgreSQL brand assets including trademarks and domain names. (link to PostgreSQL governance / link to PGCA non-profit)
  
  


• PostgreSQL developer meetings: Twice a year the developer community hosts invitation-only summits—typically coinciding with FOSDEM PGDay and PGConf.dev—to discuss “wicked problems”, technical issues, or development processes that are not easily addressed over email. Members of our Postgres contributor team often attend—and +1 to Daniel Gustafsson and Melanie Plageman for helping to organize the recent developer meetings.
  
  


• Blogging: The blog platform where our Postgres team has been publishing the most in the last year—with 48 new posts—is on Microsoft Tech Community. Blogging continues to be a super useful way to share information—and when I’m researching something at 2AM, I’m definitely appreciative of all the Postgres people who take time to share information via blog! (link to our blog on Microsoft Tech Community)
  
  


• Contribute to Patroni 3.3.x and 4.0.x: Patroni is a popular High Availability (HA) solution for Postgres. It helps you deploy, manage, and monitor HA clusters using streaming replication—and it’s open source. Alexander Kukushkin from our team is the technical lead and collaborates with engineers from different companies on Patroni. And in this timeframe there have been several new releases focused on stability, bugfixes, support for Postgres 17, and delivery of new features.
  
  I started to include individual release notes links for the last 12 months but there were too many to list: 7 minor releases and 1 new major release. The Patroni team has been prolific! If you’re interested in learning more about Patroni, Polina Bungina from Zalando will be giving a talk in June at POSETTE 2025 titled, “What is Patroni, really?” (link to Patroni release notes)

Microsoft ❤️ Postgres

This post is a snapshot of the work our Postgres team at Microsoft has done over the last 12 months. There are too many contributors to name—but without their efforts, this post would not exist.

As I wrap-up, I’m struck by just how many ways this team shows up for Postgres. We run a managed Postgres service on Azure. We contribute upstream to PostgreSQL with code, architecture, reviews, testing, and extensions. We work on open source projects like Citus and Patroni. We write blog posts. We give talks. We sponsor and help organize conferences—including POSETTE, the world’s largest virtual event for Postgres. And yes, we even host a podcast.

And maybe it’s just me—but the infographic that pulls it all together? Kind of a work of art.

**********

Acknowledgements

Gratitude 😊 for reviews of this blog post and infographic (in date order) to Charles Feddersen, Varun Dhawan, Guy Bowerman, Krishnakumar (KK) Ravi, Daniel Gustafsson, Teresa Giacomini, Harmeet Singh, Thomas Munro, David Rowley, Maxim Lukiyanov, Ibrahim Halatci, Joshua Johnson, Rakesh Gujjula, Silvano Coriani, Danyal Bukhari, Akash Rao—and also Isaac Alves & My Nguyen.