Point of Presence
Definition
A point of presence is one site within an edge network: a cage or a few racks inside a data centre or an internet exchange, holding caching servers, network gear and a connection to the local providers. When a provider says it has 120 locations, it means 120 of these. Each one keeps its own cache, so a file popular in Sao Paulo is stored in Sao Paulo and is simply absent from the machine in Oslo until somebody there asks for it. Placement follows internet geography rather than population: sites go where many networks already interconnect, which is why the list reads like a list of exchange points and not a list of big cities.
Why It Matters
The distance between a visitor and the nearest one is most of the latency they experience. Light in fibre covers roughly 200 kilometres per millisecond, and real paths are indirect, so a visitor 200 kilometres from a site sees a round trip of a few milliseconds while one 5,000 kilometres away sees 70 or more before anything is processed. Coverage gaps show up as regional complaints that make no sense from head office: the site is fine in London and sluggish in Lagos, because the nearest location is in another country. Count is a weak signal on its own — capacity, peering quality and how much cache each site holds decide whether being near one actually helps.
How It Works
Anycast announcements make every location advertise the same address, and the routing between networks delivers each request to whichever announcement is nearest in network terms. The receiving site terminates the encrypted connection, looks up the requested path in its local store, and either serves it or fetches it from the origin and keeps a copy. Because each site caches independently, the first request for a file in each region is a miss and pays the full journey; providers reduce that with a shield location that all sites fetch through. Capacity is managed by evicting the least recently used objects when the disks fill, so an infrequently requested large file may be refetched every few hours rather than sitting there indefinitely. Response headers usually name the site that answered, often a three-letter airport code such as LHR or NRT.
Real-World Example
A tour operator hosts a 9 MB brochure on 99helpers at calloway-tours.99helpers.site and sends it to mailing lists in Britain and Singapore on the same day. The first British reader causes a fetch into the London location and everyone after them is served from it; the first Singapore reader causes a separate fetch into the Singapore location. Both groups end up with downloads starting in tens of milliseconds, and the operator never learns any of this happened. It shows up only as a download count in the 99helpers analytics, evenly fast in both markets.
Common Mistakes
- ✕Reading a provider's location count as a speed ranking — a hundred small sites with thin peering can be slower than thirty large ones
- ✕Assuming a new upload is present everywhere immediately — it exists nowhere until somebody in each region requests it and causes a fetch
- ✕Blaming the origin for a slow first download — a cold miss in a distant location looks identical to a slow server unless you check whether it was a HIT or a MISS
- ✕Ignoring cache capacity — a rarely requested file gets evicted, so every few hours someone pays the full fetch again no matter how near the site is
Related Terms
Edge Network
The fleet of servers a delivery provider runs in cities around the world, positioned close to visitors rather than close to your origin. It is the machinery a CDN is built from.
CDN
A content delivery network: a set of servers spread across the world that keep copies of your files and answer each visitor from somewhere close to them. It is the service; the machines it runs on are its edge network.
Latency
Latency is the delay between sending something and it arriving — usually quoted as round-trip time, the milliseconds for a packet to reach a server and come back. It is a separate thing from bandwidth, which is how much data fits through per second once the first packet lands.
Time to First Byte
Time to First Byte, or TTFB, is the gap between a browser starting a request and the first byte of the response body arriving. It measures everything that has to happen before the page can begin — not how long the page takes to finish.
Web Analytics
Measuring who visits a site and what they look at. The two methods — a script running in the visitor's browser, or counting each request as it is served — produce different numbers, and neither of them is wrong.
Put a file online in seconds
Drop in a document, an image, a page or a whole static website and share the link — free, with no build step and no server to set up.
Host a file free →