Hyd PGDays 2025

Ragesh Hajela – Fujitsu

Chiranmoy Bhattacharya – Fujitsu

As Arm CPUs become increasingly prevalent in data centers and high-performance computing, optimizing core database operations for vector performance is critical. To enable PostgreSQL to take full advantage of modern Arm architectures, we contributed the first-ever integration of Scalable Vector Extensions (SVE) into PostgreSQL. This work delivers a portable, vector-length-agnostic SIMD backend that boosts performance across a wide range of SVE-enabled Arm CPUs.

Our initial patch introduced vectorized versions of critical routines and setting the foundation for broader SIMD acceleration in PostgreSQL. First contribution introduced SVE-accelerated pg_popcount, used in visibility map operations that are fundamental to ANALYZE and VACUUM commands. This change delivered up to 7x speedup for that function on SVE-enabled platforms. Our second contribution focuses on optimizing hex encoding/decoding for bytea data types using SVE, achieving up to 10x speedup for these routines. Together, these improvements contributed upto ~11% performance gain on few queries in the TPC benchmarks.

These kernels are written using Arm SVE intrinsics and designed to scale with vector lengths from 128 to 512 bits, enabling performance portability across CPUs like AWS Graviton, A64FX, NVIDIA Grace, and Fujitsu’s upcoming MONAKA Processor (SVE2, 256-bit).

This talk will detail the motivation, design, and implementation of SVE support in PostgreSQL, discuss upstream collaboration, and present early performance metrics. We’ll also explore how this vectorization effort aligns with PostgreSQL’s long-term goals for extensibility and multi-architecture performance optimization.

Yogesh Jain | Enterprise DB

PostgreSQL is widely used across cloud, Kubernetes, and bare metal environments. But how can we monitor it efficiently across all these setups without vendor lock-in or high costs?

In this talk, we’ll explore how OpenTelemetry helps achieve consistent and scalable observability for Postgres – no matter where it runs. We’ll walk through how to collect metrics and logs, integrate them with open-source observability tools, and optimize monitoring for both performance and cost.

Expect live demos and practical insights on making PostgreSQL monitoring simple, efficient, and future-ready with OpenTelemetry.

Key Takeaways:
– How OpenTelemetry brings consistent Postgres observability across cloud, kubernetes, and bare metal
– How to collect telemetry data (metrics and logs) for effective Postgres monitoring
– How to use open-source tools (Prometheus, Loki, Grafana) to observe Postgres
– Tips to reduce observability costs while keeping full visibility
– Live demo of Postgres monitoring in cloud-native environments

Sivasankar Prasad – Ericsson

AI projects today often rely on high-dimensional vector embeddings—for example, from text, images, audio, or user behavior—that power semantic search, recommendation systems, and natural language processing pipelines. Traditionally, handling such vectors required dedicated vector databases, introducing complexity in data pipelines and system architecture.

This talk focuses on the benefits of using pgvector, a powerful PostgreSQL extension that enables native vector similarity search directly within the PostgreSQL cluster. By embedding AI capabilities into the same environment as your relational data, pgvector eliminates the need for ETL pipelines between systems, improves performance through locality, and maintains transactional integrity. We’ll cover real-world use cases, performance considerations, and how pgvector simplifies deploying AI features like semantic search, nearest neighbor queries, and hybrid filtering using structured + unstructured data. Ideal for data engineers, ML practitioners, and Postgres users looking to integrate AI functionality with minimal infrastructure overhead.

Kshetrapaldesi Tutika | Microsoft

In this talk, I will share how PostgreSQL has improved its security features over the years—from version 10 all the way to version 17. I’ll explain changes in user management, roles, and privileges, especially the update in version 15 where users no longer get the default CREATE privilege in the public schema. I’ll also talk about column-level security introduced in version 15, with real-life examples like protecting healthcare data.

I’ll then highlight the better role management in version 17, which gives more control and flexibility through role inheritance.

The session will also cover authentication and encryption improvements:
SCRAM-SHA-256 authentication from version 10
LDAP updates in version 11
GSSAPI encryption from version 12
Stronger password policies from version 13
SSL improvements in version 14
Direct TLS handshake support in version 17

To wrap up, I’ll talk about the new incremental backup feature in version 17, which helps prevent attackers from rebuilding the database without full backup access.

Nikhil Hegde

You run a few queries in development everything looks fast. But once your app goes live, you notice strange things: slow queries, deadlocks, growing table sizes, and confusing inconsistencies. Even something as simple as SELECT COUNT(*) feels unexpectedly slow. What’s going on?

In most cases, the answer lies in how PostgreSQL handles concurrency through MVCC—Multi-Version Concurrency Control. It’s a powerful system that lets PostgreSQL support high-concurrency workloads without locking readers and writers but it also introduces complexity that can catch you off guard.

This talk will walk through how MVCC actually works under the hood: how row versions are managed, how visibility is decided, and how background processes like autovacuum come into play. We’ll connect these internals to real-world issues you’ve probably faced—and explain how to solve them.

By the end of the session, you’ll know how to:
* Diagnose and resolve common performance issues caused by MVCC behavior
* Make informed choices around isolation levels and long-running transactions
* Tune autovacuum and avoid table bloat in write-heavy workloads
* Write safer, more scalable queries in concurrent environments

Whether you’re a developer, DBA, or architect, this talk will help you stop guessing and start making informed decisions about how your PostgreSQL database behaves under load.

Vikas Gupta | Amazon

Rajesh Madiwale | Amazon

This technical presentation explores real-world challenges encountered while operating PostgreSQL in high-stakes production environments. Through an in-depth analysis of performance bottlenecks, we examine how seemingly routine operations can cascade into significant system degradation. The discussion covers critical aspects of PostgreSQL that impact database performance.
We present a systematic approach to troubleshooting, backed by metrics and practical examples, demonstrating how proper configuration tuning and proactive monitoring can prevent common failure modes. The session includes case studies of production incidents, analyzing root causes from both architectural and operational perspectives. Special attention is given to query performance optimization, resource utilization patterns and novice ideas to tackle them.

– Scary Plans & Statistics out of way
– Ineffective Vacuum process
– Impacts of Idle connections
– Impacts of MultiXacts
– Memory Utilizations
– Rouge queries consuming 100% CPU
– Lock Contention

Lahari Giddi | OpenSource DB

Keerthi Seetha | OpenSource DB 

PostgreSQL is famously reliable, but what happens when you push it beyond its comfort zone? In this talk, we unpack real-world experiments and production journeys using multi-master replication (MMR) and distributed PostgreSQL setups. We’ll explore brittle scripts, unexpected race conditions, split-brain scares, and hard-won breakthroughs. Whether you’re scaling writes, chasing high availability, or just PostgreSQL curious, you’ll walk away with practical insights, cautionary tales, and a clearer sense of how this beloved database could be Multi-master and distributed.