A request from Tokyo hits a server in Singapore. Instead of going straight there, it loops through Chicago and back. It's not a bad configuration. The network is just guessing where the machine sits - and guessing wrong.
Fact
The facts: Cloudflare keeps copies of your content in a tiered layer and measures response time to decide which such data center sits closest to your server. Behind a shared address on a big cloud that measurement lies, because the address is the same from everywhere while the machine sits in one specific region. On 10 July 2026 the company expanded a feature introduced in April: instead of the network guessing, you name the region directly, for example aws:us-east-1, and Cloudflare maps it to the right tiered data center and keeps a fallback in case it drops. It works for AWS, Google Cloud, Azure and Oracle Cloud, configured through the dashboard, API and Terraform, available on every plan including the free one, at no extra cost.
Let's not dress it up. This isn't a revolution. It's a fix for something that should have worked from the start. The automation that supposedly knew where your server was, didn't. Instead of making it smarter, they added a field where the human just tells the truth. Sometimes the best solution is to ask the one who knows.
It matters to anyone who keeps their server on a big cloud and runs their site through Cloudflare. Hundreds of milliseconds on a round trip sound small, until they add up to a slow site. And since it's available on the free plan, small teams can use it too. That's not the usual case.
Official primary sources
Cloudflare BlogThe visual is generated code-art, no third-party images.