Newsdesk
Engineering
Data & AI
Industries
Enterprise Systems
Go-to-Market
Longform
DesignIndiaAll stories

Databases & Data

06 October 2026

63 new posts across 9 database and data engineering sources — 7 worth your time.

One week of Databases & Data, 7 stories, as published.

Performance

Alexander Ioffe: I ran the same query twice and Postgres disagreed with itself

The author investigates why two identical EXPLAIN ANALYZE runs on statistically identical data produced opposite query plans (73 vs 3,449 heap blocks) with a tiny 1.5% margin. The post uses ExoBench to measure the planner's behavior and demonstrates an index shape that eliminates this unpredictability.

Why it matters — A tiny statistical margin can cause the Postgres query planner to fluctuate wildly between plans, but specific index shapes can eliminate this unpredictability.

Planet PostgreSQL · 06 October 2026 · Read the original →

Postgres

Hollis Varden: UUIDv4 or UUIDv7 for a primary key? 5 million rows on PostgreSQL 18

This post benchmarks UUIDv4 against the newly native UUIDv7 primary keys in PostgreSQL 18 using a 5-million-row dataset. The results show that UUIDv7 inserts took 27 seconds compared to 46 seconds for UUIDv4, with the UUIDv4 index being 29% larger, while random lookups performed identically.

Why it matters — Sequential UUIDv7 primary keys can reduce insert times by over 40% and index size by nearly 30% compared to random UUIDv4, without sacrificing lookup performance.

Planet PostgreSQL · 06 October 2026 · Read the original →

Performance

Chris van Eijk: Row-level security performance in PostgreSQL, measured

Evaluates the performance overhead of Row-Level Security (RLS) policies in PostgreSQL 17. It analyzes the specific costs associated with function volatility, the use of `(SELECT ...)` wrappers, membership subqueries, and how the leakproof rule can inadvertently disable indexes.

Why it matters — RLS policies can silently disable index scans if the underlying functions are not marked leakproof, severely degrading query performance.

Planet PostgreSQL · 06 October 2026 · Read the original →

Analytics & OLAP

How LinkedIn extended ClickHouse from distributed tracing to metric discovery and analytics

Details how LinkedIn scaled its ClickHouse-based observability stack to consolidate over 13 billion metrics into a single index. The system successfully handles more than 150,000 queries per minute, expanding ClickHouse's role from tracing to high-throughput metric discovery.

Why it matters — Consolidating massive-scale tracing and metric data into a unified ClickHouse index can support extremely high query volumes (150k+ QPM) while simplifying observability architecture.

ClickHouse · 06 October 2026 · Read the original →

Performance

Handling hot shards

Explores the challenges of sharding database tables on a `tenant_id` column in production. It discusses how Neki's router manages data topology to handle uneven write and read distributions across shards when dealing with hot tenants.

Why it matters — Sharding by tenant ID inevitably creates hot shards in production, requiring routing layers that can dynamically manage uneven data topology and query distribution.

PlanetScale · 06 October 2026 · Read the original →

Performance

Designing Neki for performance

This post details the design of Neki, a routing layer positioned between PostgreSQL clients and servers to enable sharding. It addresses the performance challenge of introducing an extra network hop and discusses the trade-offs of making multiple database servers appear as a single instance.

Why it matters — Adding a routing layer for database sharding introduces latency overhead; optimizing this layer requires minimizing processing time to offset the unavoidable extra network hop.

PlanetScale · 06 October 2026 · Read the original →

Reliability

How we systematically improved our reliability

Following two major incidents in a single week, Neon outlines the systematic engineering efforts implemented to improve platform reliability. The post details the specific technical and operational changes rolled out to maintain high uptime across a rapidly growing fleet and customer base.

Why it matters — Post-incident reliability engineering requires a systematic, fleet-wide approach to infrastructure hardening rather than isolated quick fixes.

Neon · 06 October 2026 · Read the original →

Another week of fighting the database we chose to make our lives easier.

7 stories, every Tuesday

Published here every week. Follow by RSS to get it as it lands.

← Previous issue Next issue →