10 November 2026
83 new posts across 10 database and data engineering sources — 9 worth your time.
Nikolay Samokhvalov: UUIDv4 vs UUIDv7: what a random primary key costs in Postgres
This post benchmarks UUIDv4, UUIDv7, and ordered primary keys on Postgres 18 using a 100M row dataset. It quantifies the performance penalty of random keys, showing UUIDv4 inserts at roughly 65% the rate of ordered keys while generating up to 1.9x the WAL volume per row.
Planet PostgreSQL · 10 November 2026 · Read the original →
Chris van Eijk: json vs jsonb in PostgreSQL: size and speed, measured
A detailed comparison of json and jsonb storage in Postgres 17 and 18 across six different document structures. The results show that while jsonb is often faster to query, it can be up to 41% larger than raw json depending on the document shape.
Planet PostgreSQL · 10 November 2026 · Read the original →
Fencing a Postgres primary that might still be alive
This technical deep dive explains the 'fencing' problem in database failover, where a promoted standby must ensure the old primary cannot accept writes. It details specific implementation strategies for bare metal, AWS, and GCP, and explains the rationale behind a fail-open design.
ClickHouse · 10 November 2026 · Read the original →
Lakebase Postgres is built on object storage, but 99.99999% of reads never touch it
The post examines the architecture of running Postgres on top of object storage like S3. It explains how a multi-tiered caching strategy allows 99.99999% of reads to be served from local memory or NVMe, bypassing the high latency of the object store.
Neon · 10 November 2026 · Read the original →
The hidden latency inside your database
An analysis of the 'hidden' network latency that exists inside cloud-managed database clusters, specifically between the compute engine and network-attached storage (NAS). It argues that this internal hop is often a larger bottleneck than the distance between the application and the database.
PlanetScale · 10 November 2026 · Read the original →
Richard Yen: An Index Scan Is Only Half the Story
A migration case study where a query's performance degraded from tens of milliseconds to over a second despite identical schemas and indexes. It explores how infrastructure environment differences and query planner behavior can diverge during cloud migrations.
Planet PostgreSQL · 10 November 2026 · Read the original →
What I learned from a month with WalShadow on ClickHouse Managed Postgres
A report on a month-long test of WalShadow for Postgres-to-ClickHouse replication. It provides a direct comparison of replication latency against standard pipes and evaluates how the system handles schema changes and data recovery.
ClickHouse · 10 November 2026 · Read the original →
Snowflake Delivers Autonomous Data Engineering
Snowflake announced new autonomous data engineering and real-time analytics innovations at Expedition 2026. These capabilities leverage AI agents to automate recurring data pipeline fixes, enabling end-to-end real-time analytics with a median data-readiness latency of 3 seconds for dashboards, applications, and AI agents.
Snowflake · 10 November 2026 · Read the original →
Confluent Introduces Streamhouse Architecture
Confluent introduced its Streamhouse architecture, a new data architecture designed to provide real-time business context for AI agents and applications. With Lightning Tables now generally available on AWS, the platform offers low-latency SQL and API access to live streaming data, with changes available within hundreds of milliseconds and queries returning in 10-100 ms at over 1,000 queries per second.
confluent.io · 10 November 2026 · Read the original →
Still micro-optimizing Postgres while the vendors invent new words for databases.
9 stories, every Tuesday
Published here every week. Follow by RSS to get it as it lands.