Best of Database — September 2025
- 1
- 2
Kiki's Bytes·36w
How Uber Saved $6,000,000 in Database Costs
Uber reduced database costs by $6 million annually by migrating their financial transaction system from DynamoDB to DocStore, their custom MySQL-based distributed database. The migration involved building LedgerStore 2.0 with immutable records, data integrity guarantees, and solving hot partition problems through deterministic sharding. The process included careful data backfilling, shadow writes for real-time sync, and dual read validation to ensure zero downtime during the transition.
- 3
Dev Tools·35w
Free Valentina Studio 16 now available
Valentina Studio 16 has been released as a free database management tool supporting major databases including PostgreSQL, MongoDB, MySQL/MariaDB, MS SQL Server, SQLite, DuckDB, and ValentinaDB. The PRO version includes advanced diagramming, reporting, and forms creation capabilities similar to MS Access or FileMaker. A 5-connection version of Valentina Server 16 is also available for free.
- 4
JetBrains·35w
A Farewell to Consoles
DataGrip 2025.3 replaces query consoles with query files to improve workflow consistency. The change addresses longstanding issues where consoles lived outside project structure, were hard to find, and had limited context switching capabilities. Query files are now stored in a visible project folder, support full file operations, and maintain database context with improved visual indicators. Existing consoles will be migrated automatically with a one-release transition period.
- 5
Crunchy Data·36w
Get Excited About Postgres 18
PostgreSQL 18 introduces significant performance improvements including asynchronous I/O for faster reads, UUID v7 support for better indexing, B-tree skip scans for multi-column indexes, virtual generated columns as the default, and OAuth 2.0 authentication support. The async I/O feature allows workers to batch reads and optimize idle time, while UUID v7 provides better locality through timestamp-based prefixes. Skip scans enable using multi-column indexes even when leading columns aren't in query conditions, and virtual generated columns compute values on-the-fly without storage overhead.
- 6
Collections·34wKey Features and Enhancements in PostgreSQL 18
PostgreSQL 18 introduces major performance improvements including asynchronous I/O that delivers up to 3x faster storage reads, NUMA support, and IO_uring on Linux. The release adds enterprise features like OAuth 2.0 authentication, temporal constraints, and virtual computed columns. Key enhancements include better extension loading for Kubernetes deployments, improved logical replication conflict handling, UUIDv7 support, and advanced vacuum operations for large tables. The update also brings query planning improvements with skip scan lookups and enhanced Unicode collation support.
- 7
Architecture Weekly·36w
Just use SQL they say... Or how accidental complexity piles on
Explores how the seemingly simple approach of "just use SQL" can lead to accidental complexity as requirements grow. Starting with basic SQL queries in TypeScript, the article demonstrates how adding features like filtering, parameterization for SQL injection prevention, and dynamic query building quickly escalates into building a custom query compiler. The piece argues that while "just" solutions work initially, they often evolve into the same complexity that existing tools were designed to solve, emphasizing the importance of considering long-term needs when choosing between DIY and established solutions.
- 8
Hacker News·36w
OrioleDB Patent: now freely available to the Postgres community
Supabase has completed the acquisition of OrioleDB and is making the related patent freely available to all users. OrioleDB is a high-performance storage extension for PostgreSQL that shows 5.5x performance improvements over traditional heap storage. The project remains open source with an open contribution model, and Supabase aims to eventually upstream OrioleDB into PostgreSQL itself. The patent license is intended as defensive protection for the open source community rather than offensive enforcement.
- 9
DuckDB·35w
Announcing DuckDB 1.4.0
DuckDB 1.4.0 'Andium' introduces Long Term Support with 1 year community maintenance, database encryption using AES-256, MERGE statement for upsert operations, Iceberg write support, CLI progress bar with ETA, FILL window function for interpolation, and performance improvements including sorting rework and materialized CTEs. The release also includes macOS notarization and moves Python integration to a separate repository.
- 10
Postgres 18: OLD and NEW Rows in the RETURNING...
PostgreSQL 18 introduces the ability to access both OLD and NEW row values in RETURNING clauses for INSERT, UPDATE, DELETE, and MERGE operations. This enhancement allows developers to see previous and current data states in a single query, making it easier to detect whether an upsert operation inserted a new row or updated an existing one without relying on internal implementation details like xmax values.
- 11
Awesome·37w
Google just dropped a new language for deductive database programming...
Google released Mangle, a new programming language for deductive database programming built on Datalog. It enables querying scattered, inconsistent data from multiple sources (databases, APIs, logs) using a single logic layer with declarative rules and facts. Unlike traditional query languages, Mangle includes a reasoning engine that can infer relationships and handle complex recursive operations like dependency tracing and compiler analysis without manual step-by-step coding.
- 12
PlanetScale·34w
PlanetScale for Postgres is now GA — PlanetScale
PlanetScale has officially launched their Postgres database service, moving from private preview to general availability. The managed service combines PlanetScale's proven infrastructure with Postgres, targeting companies that need scalable database solutions. Several companies including Convex, Supermemory, and Layers have already migrated to the platform. PlanetScale is also developing Neki, a Postgres sharding solution built from first principles that will eventually be open-sourced for demanding workloads.
- 13
Vercel·34w
Claimed deployments now include third-party resources
Vercel introduces a new feature allowing third-party platforms to create deployments with attached resources that users can later claim and take full ownership of. The feature enables instant deployment of complete application stacks, including databases and other services, through a two-step process: third-parties create and deploy projects via the Vercel API, then users claim ownership of both the deployment and its resources. Prisma is the first marketplace provider to support this functionality, allowing customers to deploy databases and Vercel-hosted applications as bundled stacks.
- 14
P99 Conf·34w
Books by P99 CONF Speakers: AI Engineering, Latency, Distribtuted Systems & More
A curated collection of technical books authored by P99 CONF speakers covering AI engineering, distributed systems, database performance optimization, latency reduction, and PostgreSQL query optimization. The books range from foundational concepts to advanced implementation techniques, with special discounts available through conference sponsors O'Reilly and Manning Publications.
- 15
- 16
- 17
Web Dev Simplified·35w
How To Handle Data Access Like a Senior Dev
Demonstrates how to implement a proper data access layer (DAL) in Next.js applications to centralize database operations, user authentication checks, and error handling. Shows the evolution from naive direct database access in components to a sophisticated system that returns either success or error states, enabling flexible handling of different scenarios like API routes vs page rendering. Covers both basic and advanced DAL patterns with TypeScript for better type safety and maintainability.
- 18
depesz·35w
Waiting for PostgreSQL 19 – Add date and timestamp variants of random(min, max). – select * from depesz;
PostgreSQL 19 introduces new random() function variants for generating random dates and timestamps within specified ranges. The update adds three new functions: random(min date, max date), random(min timestamp, max timestamp), and random(min timestamptz, max timestamptz). These functions simplify test data generation by eliminating the need for manual calculations with intervals and mathematical operations.
- 19
pgEdge goes Open Source
pgEdge has relicensed all core components of their Distributed Postgres platform from a proprietary pgEdge Community License to the permissive PostgreSQL License. This includes their replication engine Spock and extensions like Snowflake and Lolor. The change makes pgEdge's distributed PostgreSQL technology fully open source, allowing unrestricted use and modification of the source code.
- 20
HelixML·34w
Kodit 0.5: All Things Git
Kodit 0.5 introduces a major architectural shift from directory-based to Git-based domain modeling, enabling incremental indexing and better repository version handling. The release includes LiteLLM integration for multiple embedding providers, a refactored queue-based indexing pipeline with status endpoints, and database performance improvements. Breaking changes include database schema restructuring and removal of auto-indexing commands. Future plans include repository-specific enrichments for users, developers, and readers, plus a user interface and expanded indexing capabilities beyond code.
- 21
MotherDuck·37w
Announcing Pg_duckdb Version 1.0
Pg_duckdb version 1.0 is now available, bringing DuckDB's vectorized analytical engine directly into PostgreSQL as an extension. This integration enables faster analytical queries on PostgreSQL data without requiring separate data warehouses or complex ETL processes. The extension allows querying PostgreSQL tables with DuckDB's performance benefits, accessing external data lake files (Parquet, CSV, JSON), and joining local PostgreSQL data with remote cloud storage files in single queries. Performance improvements show up to 4x speedup with indexes and dramatic improvements for queries that previously timed out. The release includes enhanced MotherDuck integration for serverless analytics scaling.
- 22
freeCodeCamp·35w
A Brief Introduction to SQLite
SQLite is an underappreciated embedded database that runs directly in applications without server setup or configuration. It powers smartphones, browsers, and desktop apps worldwide. This tutorial covers SQLite integration through practical examples in C/C++, Python, and Java, including database design, low-level API usage, Flask web development, and JDBC implementation. The content emphasizes SQLite's simplicity and power for local data storage without the complexity of full database servers.
- 23
PostgreSQL·36w
PostgreSQL: pgexporter 0.7
pgexporter version 0.7.0 has been released with improvements to core metrics and new features. This Prometheus exporter for PostgreSQL includes extension support developed as part of a Google Summer of Code project. The tool helps monitor PostgreSQL databases by exporting metrics to Prometheus.
- 24
ClickHouse·33w
How we scaled raw GROUP BY to 100 B+ rows in under a second
ClickHouse Cloud introduces parallel replicas, enabling GROUP BY queries to scale across thousands of cores automatically. The feature processes 100 billion rows in under half a second without pre-aggregation or data reshuffling. By distributing work across multiple nodes using partial aggregation states, queries achieve terabyte-per-second throughput with linear scaling. The system uses granule-based load balancing and safeguards to optimize performance, making analytical queries interactive even at massive scale.
- 25
The Daily WTF·34w
One Last ID
A web development company made a costly SQL mistake by adding an unnecessary FROM clause to a last_insert_id() query, causing MySQL to return millions of duplicate values instead of a single ID. This simple error resulted in 30MB responses for each INSERT operation, generating terabytes of unexpected data transfer costs and blowing the operational budget. The technical fix was simple, but coordinating the deployment took six weeks, including two weeks just to convince the company the problem existed.
