Measured August 11, 2026 · smallest server target
This is the public Amber V2 site: controllers, ECR rendering, layout, content, and 26,271 bytes over HTTP. It ran on the DigitalOcean Basic one-vCPU, 512 MB class—the $4/month size when the lab ran.
What the number means
A median of 5,906.72 requests per second means the middle repeated run completed that many full homepage responses each second under the test's connection and keep-alive settings. It does not mean every database-backed page will do the same thing.
The practical result is room: a documentation site, small product, or portfolio of read-heavy applications can begin on modest infrastructure without treating performance as an afterthought.
5,906.72 req/s median
21,795 req/s median
19,488 req/s median
Read this before repeating the number
No database was involved. A query, remote API, file store, authentication flow, or application code can become the limiting factor.
The 1,000 WebSocket clients were idle after joining. The result proves connection holding and concurrent HTTP work, not 1,000 clients broadcasting continuously.
Sequential shared-vCPU stages were noisy. They are published as observed evidence, not presented as a causal scaling curve.
This is not an SLA or cross-framework ranking. Hardware, proxies, TLS, traffic shape, application behavior, and deployment tuning will change the result.
Reproduce or audit it
The raw records preserve exact stages and measurements for reviewers and tools. They support this page; they are not the page.