Urgent.News

What's breaking now, across thousands of outlets.

Sunday, 8 November 2026 Reading Room

Tech

Cloudflare Scales D1 Control Plane to Millions of Databases

In a recent article, Cloudflare shared the redesign of the control plane of its D1 serverless SQL database to support millions of databases, avoiding centralized bottlenecks, implementing retries and idempotency, and using eventual consistency. By Renato Losio

infoq.com ·

Cloudflare has recently announced that it has redesigned the control plane of its D1 serverless SQL database to support millions of databases, addressing scalability issues and avoiding centralized bottlenecks. The new architecture is built around Postgres for transactional metadata, ClickHouse for analytics, and Workers KV for globally replicated routing data.

By keeping query processing at the edge, separate from centralized management operations, Cloudflare has implemented distributed metadata caching, connection pooling, asynchronous workflows, and periodic health checks to improve scalability and recover from failures. The new system uses 30-second heartbeats to reconcile each D1 Durable Object with Postgres, the source of truth for database existence and account limits, ensuring self-correction of failed creation or deletion operations.

Brief written by urgent.news from InfoQ's own syndicated text. Machine-written — may contain errors; check the original before relying on it.

Read the original at infoq.com →

More in Tech

I made a free, open source alternative to Brilliant (140+ interactive math and science courses)

What a Free, Open-Source Brilliant Clone Teaches Us About Lazy Eye Training Someone on Lemmy posted a free, open-source alternative to Brilliant this week: 140+ interactive courses, 1000+ lessons in…

  • OpenAha project offers 140+ interactive math and science courses
  • Feedback loop approach encourages user engagement and contributions
  • AmblyoPunch shares similar adaptive difficulty system for vision training

·dev.to

It won on the steady sellers and lost on the rest

On one Walmart store, a machine-learning forecast was 3.9 percent worse than a 28-day moving average on absolute error, and the worse forecast for 58 percent of the 823 items.

  • Walmart's machine-learning forecast performed 3.9% worse than 28-day moving average on average error
  • Model excelled on steady sellers but struggled with intermittent and lumpy items
  • Moving average outperformed model for 77% of the store's items

·dev.to

How does FORECAST.ETS predict the December sales peak?

FORECAST.ETS(...) looks like magic: give it three years of monthly sales and it sends back a December higher than all the other months, a January that drops and a confidence interval around both.

  • FORECAST.ETS predicts December sales peak
  • 36-month monthly sales table input
  • Predicted value for December 2026 is 20.654

·dev.to

How we check every hour that our PDF API stores nothing

Every hour, a script sends a document with a unique marker through our public API, then searches our logs, temp files, containers and database for it.

  • Hourly script tests PDF API for no document storage
  • Unique marker placed in HTML document for detection
  • Test results published publicly for transparency

·dev.to

Serving Sage-A1.5 on an RTX 4090: one broken flag, one queue, and a self-check mode to measure

Run Sage-A1.5 as a research agent on your own card and the surprises are not in the model. They hide in a reasoning flag, a request limit, a token budget and one system prompt string.

  • Sage-A1.5 runs efficiently on a single RTX 4090 with compressed 22 GB GGUF file
  • Reasoning flag and sub-agents may cause compatibility issues with llama.cpp
  • Verification mode improves accuracy on BrowseComp benchmark from 72.5% to 82.9%

·dev.to

Auditing the Gaps Around Recurly's Own Dunning

Recurly ships a serious dunning system. Automated retries, account updater integrations, and configurable campaigns handle the mechanical core of failed-payment recovery competently, and teams…

·dev.to

More from Sunday 8 November →