Emplacements périphériques de CDN et points de présence

A Point of Presence (PoP) is a physical data center location where a CDN places its edge servers. Each PoP caches copies of your content so requests can be served from a nearby location instead of your distant origin server.

• PoPs are spread across cities and regions worldwide
• Each PoP contains one or more edge servers
• The goal is to shorten the network path between user and content

An edge location is the specific server within a PoP that actually stores and serves cached content. When a user requests an asset, the CDN routes them to the closest healthy edge location.

This reduces round-trip time (RTT) and lowers latency dramatically compared to fetching from a single origin.

Data travels at finite speed. The farther a user is from your origin, the longer each request takes. By placing edge locations near users, CDNs cut down the physical distance packets travel.

• Lower latency means faster page loads
• Fewer network hops means fewer chances for packet loss
• Local delivery reduces load on your origin

Most CDNs use Anycast routing. The same IP address is advertised from many PoPs, and the network naturally routes each user to the nearest one.

This makes failover seamless: if a PoP goes down, traffic shifts to the next closest location without any DNS change needed.

Some CDNs add a second tier called a regional edge cache between the small edge PoPs and the origin. Less popular content evicted from an edge can still be served from the larger regional cache instead of hitting the origin.

• Edge PoP: smallest, closest to users
• Regional edge cache: larger, serves multiple PoPs
• Origin: your source of truth

A typical request path looks like this:

• User requests an asset
• DNS / Anycast routes to nearest edge PoP
• Edge serves a cache hit, or fetches from regional cache
• Regional cache serves it, or pulls from origin
• Content is cached at each tier on the way back

You can roughly gauge which PoP serves you by inspecting response headers. Many CDNs expose the serving PoP in a header so you can confirm users hit a nearby location.

Not every edge location stores every file. Edges hold a limited cache, so they keep the most frequently requested content. Rarely accessed files may be fetched on demand from a higher tier.

This is why the first request in a region can be slower (a cache miss) while later ones are fast.

When choosing a CDN, coverage matters. A provider with more PoPs in your users region delivers lower latency there. Evaluate the PoP map against where your audience actually lives.

• Global apps benefit from broad coverage
• Region-focused apps benefit from dense local PoPs

Because edge locations absorb most traffic, your origin sees only a fraction of requests (cache misses and revalidations). This offloading protects your origin during traffic spikes and lowers bandwidth cost.

Key points to remember about edge distribution:

• PoPs are physical locations; edge servers within them cache content
• Anycast routes users to the nearest PoP automatically
• Multi-tier caching (edge + regional) limits origin hits
• More relevant PoPs equals lower latency for your users

Test your understanding of edge locations and PoPs.

You learned how CDNs use Points of Presence and edge locations to serve content close to users. You saw how Anycast routing picks the nearest PoP, how regional edge caches add a second tier, and how this offloads your origin and cuts latency.