The Booster Blog

Performance experiments, engineering deep-dives, and customer stories.

performancecase study

Cutting LCP from 4.1s to 1.8s on a 12-year-old Shopify store

How we used the Booster edge layer, image rewriting, and a script-loader rebuild to halve LCP on a legacy storefront with 280 third-party scripts. The unexpected villain wasn't JS — it was a poorly configured font face.

engineering

Why we shipped a custom Brotli dictionary for JSON APIs

Stock Brotli is great at HTML and CSS. For JSON-shaped payloads, a 4 KB tuned dictionary buys another 7-12% — and it's free if your API response shapes are stable. Here's how we built and deployed ours.

infra

Inside Booster's request router: 8μs of joy

The router that fans out incoming HTTPS at our edge has a budget of 8 microseconds before it must hand off. We rewrote it from a coroutine-per-request model to a lock-free dispatch table. This is the postmortem of the migration.

product

Synthetic checks: the underrated half of performance monitoring

RUM tells you what real users saw. Synthetic checks tell you what your site does on demand, at a fixed time, from a fixed network. Run both, or you'll get blindsided.

case study

From AWS CloudFront to Booster: a SaaS migration in 2 hours

A B2B SaaS doing $9M ARR migrated their entire production stack off CloudFront in a single afternoon. Here's the runbook they used and the three gotchas they hit.