Cloudflare’s Network Doubles CPU Capacity and Expands Into Ten New Cities in Four New Countries



People are saying that the cloudflare edge requests the origin using the edge closest to user instead of the edge closest to the origin is that true?

For example, if I had a server in Europe, and a user in south america tried to connect to me, what would be the route used ? don’t need specifics just a general understanding

The user connects to the closest Cloudflare POP ( as determined by BGP, so might not be the physically closest). That POP connects back to your Origin server, essentially via the public Internet.

Some options, such as tiered caching and Argo, change that default behaviour.


I still don’t understand.

So a user in south america connecting to a origin in europe would be like this :

user > closest edge > origin server > user ?

This doesn’t seem to be the best routing option for me. Shouldn’t cloudflare use the closest edge to the ORIGIN to pull the requests instead ?

Users don’t connect to Origins. User makes a request to the closest Cloudflare Edge. If that server has a cached asset it responds with that asset. If it does not have the asset, that CF Edge makes a request to the Origin, and then responds to the user.

For a heavily cached application, then most requests are served from a server physically close to a user, which minimises the latency.

If the user and Origin are very far away from
each other the transmission latency is a function of the distance between them, I don’t think there is a massive saving in what you propose. I don’t have specific data, but my instinct says that moving the latency to the User-Edge connection would result in very poor performance.

1 Like

This topic was automatically closed 15 days after the last reply. New replies are no longer allowed.