Technical
Point of presence
Also called PoP.
Point of presence is a physical site where a CDN or network provider keeps servers close to users, usually inside a data centre in a major city. A provider runs dozens to hundreds of them.
How it is measured
Many CDNs reveal the PoP in a header: a `cf-ray` value ending in an airport code such as SIN, or an `x-served-by` naming a site like fra. Run `curl -sI` from several networks and note the code along with the cache status.
Distance from the visitor to the PoP is the number that matters. Time a ping or the TLS handshake. 10 to 30 ms means a nearby PoP, and 150 ms or more means you are being served from another continent.
Worked example
A site's visitors in Lagos see a time to first byte of 620 ms. The `cf-ray` suffix is LHR, so traffic from Lagos exits to London because no closer PoP sits on the local ISP's route. 480 of 500 slow samples carry that tag.
Once the provider lights up a Lagos site and the ISP peers there, the suffix changes to LOS and the time to first byte falls to 140 ms.
How it differs
A point of presence is a place. A CDN is the network of such places plus the software that caches and routes. A PoP may hold a cache, but a CDN's reach is the sum of all its PoPs and how traffic gets to them.
Common errors
Assuming a PoP count equals coverage. Treating a PoP in the same country as nearby. Ignoring cold caches at a rarely used PoP. Judging from one test location. Reading the PoP code as the visitor's location.
In practice
Probe from your real visitors' regions, record which PoP answers, and ask your provider about the slow places. Compare cache-hit ratio per PoP rather than only in total.