OffNet Newsroom

Daily topic roundup

Database Technology

Saturday, August 08, 2026 · 7 stories, curated & summarized — click any story for the source.

Hacker News (100+ points) general

Postgres analytics speedup via batching, operator fusion, and SIMD

A technical deep-dive demonstrates how combining batch processing, operator fusion, and SIMD instructions can accelerate Postgres analytical queries by up to 300x. The approach optimizes the query engine by reducing row-by-row overhead and leveraging modern CPU vectorization capabilities. This represents a significant architectural shift for handling large-scale analytical workloads within the Postgres ecosystem.

  • Operator fusion reduces intermediate result materialization, lowering memory pressure and CPU cycles.
  • SIMD vectorization processes multiple rows per instruction, drastically improving throughput for analytical aggregates.
  • Batching transforms row-at-a-time execution into bulk operations, optimizing cache locality.
  • These techniques are critical for moving heavy analytical loads closer to raw hardware performance limits.
  • Implementation details suggest potential for custom extensions or forked engines to exploit these optimizations.

Shaun Thomas concludes his series on PostgreSQL 19 by highlighting the release's extensive scope, noting it contains more enhancements than previous versions. The post serves as a final overview of new syntax features and other marquee additions before the official launch. It emphasizes the importance of reviewing detailed changelogs to catch subtle but significant updates that might otherwise be missed.

  • Postgres 19 is positioned as a major release with an unusually high volume of enhancements.
  • Reviewing detailed release notes is critical to identify subtle syntax and feature updates.
  • The author's series covers multiple marquee features beyond just syntax changes.
  • This post acts as a final roundup of key improvements in the upcoming version.
WORTH QUOTINGPostgres 19 Release OverviewPostgres 19 contains more enhancements than previousversions.— Shaun Thomas
GitHub Trending (daily) githubrepos ⚠ unverified date/source

GitHub: celld offers self-hosted distributed Durable Objects using SQLite and S3

Celld is an open-source daemon that enables self-hosting Cloudflare Workers and Durable Objects on private infrastructure. It addresses the limitations of shared databases by treating each object as an isolated SQLite database, with state replicated to an S3-compatible bucket for coordination. Nodes operate without a central control plane or consensus mechanism, relying instead on the object store for synchronization.

  • Each Durable Object is its own SQLite database, eliminating contention and blast-radius failures of shared DBs.
  • Nodes coordinate solely through an S3-compatible bucket, removing the need for a control plane or consensus layer.
  • Idle cells hibernate to near-zero resource usage, optimizing cost and efficiency.
  • Fleet shares deployment and state via S3, allowing decentralized scaling without complex infrastructure.
HOW IT WORKSCelld Decentralized Sync Flow1Isolated SQLite DB per Object2State Replication to S3 Bucket3S3-Based Coordination Between Nodes4Idle Cells Hibernate to Zero

Percona evaluated their DuckDB-based MySQL storage engine using a 500GB TPC-H dataset comprising three billion lineitem rows. The test ran on an 80-core server with 187GB of RAM, comparing InnoDB, plain DuckDB, and the new MySQL+DuckDB hybrid engine side-by-side. This scale factor provides a concrete view of how the columnar storage layer performs under significant data volume constraints.

  • Validates DuckDB engine capacity at 3 billion rows, relevant for mid-sized analytics workloads.
  • Runs on commodity hardware specs (80 cores, 187GB RAM) rather than specialized big data clusters.
  • Enables columnar processing directly within MySQL without requiring external data warehouse pipelines.
  • Performance baseline established against InnoDB and pure DuckDB for hybrid query optimization.

Percona has introduced Percona Search for MongoDB, a downstream distribution of mongot, the engine behind MongoDB Atlas Search. This release provides full-text and vector search capabilities to self-managed MongoDB deployments. It enables the use of advanced AI-driven search techniques without requiring a cloud provider.

  • Percona Search is a downstream build of mongot, identical to the engine in Atlas Search.
  • Enables full-text and vector search capabilities for self-hosted MongoDB instances.
  • Supports AI application development using advanced search techniques on-prem or in private clouds.
  • Offers a path to avoid vendor lock-in associated with managed MongoDB Atlas services.
TRADE-OFFAtlas vs Percona SearchMongoDB AtlasManaged cloud serviceVendor lock-in riskIncludes mongot enginePercona SearchSelf-managed deploymentAvoids vendor lock-inIdentical mongot enginevs
PostgreSQL News database ↺ since 08-04

pgBackRest 2.59.0 Released with Enhanced Backup and Security Features

The pgBackRest community has released version 2.59.0, continuing its reputation as a scalable backup and restore solution for PostgreSQL. This update introduces malware and ransomware protection capabilities alongside block incremental backups to optimize storage and performance. The release maintains support for parallel operations, multiple compression types, and encryption to ensure robust data safety across diverse infrastructure.

  • Block incremental backups reduce storage overhead and speed up recovery windows.
  • New malware/ransomware protection adds a critical layer of data integrity assurance.
  • Parallel async archiving and resume capabilities minimize downtime during backups.
  • Support for S3, Azure, GCS, and SFTP ensures flexible repository options.
  • Page checksums validation during backup helps detect corruption early.
HOW IT WORKSpgBackRest 2.59.0 Backup Pipeline1Parallel async archiving starts2Block incremental backups run3Malware and ransomware scan4Resumable archive completes
PostgreSQL News database ↺ since 08-05

pgAdmin 4 v9.17 Released with 7 CVE Fixes and Helm Updates

The pgAdmin 4 Development Team has released version 9.17, addressing seven security vulnerabilities including CVE-2026-17346 through CVE-2026-17351 and CVE-2026-17566. The update introduces new features such as including authenticated user identity in HTTP access logs and adding an opt-in Gateway API HTTPRoute template to the Helm chart. Additionally, a new preference allows capping the row count fetched by the plain View Data action to improve usability on large tables.

  • Patch immediately to resolve 7 critical security vulnerabilities (CVEs).
  • Update Helm charts to support opt-in Gateway API HTTPRoute templates.
  • Configure row count caps in View Data to prevent performance issues on large tables.
  • Enable authenticated user identity logging for improved access auditing.
CHECKLISTpgAdmin 4 v9.17 Patch GuidePatch immediately for 7 CVE fixesUpdate Helm charts for Gateway APICap row counts in View DataEnable user identity in access logs