Content delivery network

A CDN works by placing cache servers between users and an origin server. When a browser requests an image, stylesheet, video segment, or other resource, the CDN uses DNS, anycast routing, or an application-level request path to select an appropriate point of presence. If the object is cached there, the edge returns it immediately; otherwise, the edge retrieves it from the origin, stores it according to cache rules, and forwards it to the user. This arrangement reduces the distance that frequently requested objects travel and limits repeated origin fetches.¹

Cacheability depends on the response's headers, URL, method, and content type. Cache-Control, Expires, and validation mechanisms such as ETag help determine whether an object remains fresh or must be revalidated. Dynamic or personalized responses are commonly passed through rather than stored, although some CDNs can execute logic or cache selected fragments at the edge.

CDNs improve performance chiefly by shortening network paths and serving static assets over connections optimized for many simultaneous users. They can also maintain service during a regional failure by routing traffic to another point of presence, although a CDN does not remove the need for a resilient origin, authoritative DNS, and sound application design. Internet exchange points and direct interconnection with access providers can further reduce congestion on the path to users.²

Many CDNs provide transport encryption, TLS certificate management, web application firewalls, bot controls, and distributed denial-of-service mitigation. These functions make a CDN part of an organization's security perimeter, not merely a faster file cache. Operators must still protect origin addresses, keep certificates and software current, and avoid treating traffic filtering as a substitute for application-layer security controls.³

Cache policy determines whether a CDN accelerates a site or serves incorrect content. Long-lived immutable assets, often identified with versioned filenames or content hashes, can be cached safely for extended periods. Frequently changing pages require short lifetimes, revalidation, or explicit purges. A cache miss, an expired object, or a poorly chosen query-string policy can send requests back to the origin and create a sudden load spike known as a cache stampede.

CDN configuration also involves cost, privacy, and observability. Serving traffic from many regions may lower user latency while increasing transfer fees or complicating data-residency obligations. Logs and request metadata can reveal sensitive information, so retention and access policies matter. Standards such as HTTP and its caching directives provide the common language, but provider-specific rules govern invalidation, routing, compression, image transformation, and edge computing.⁴

CDNs are not limited to web pages: they commonly distribute software updates, game assets, live and on-demand video, map tiles, large scientific datasets, and Internet-of-Things firmware. Video delivery often uses adaptive bitrate streaming, dividing media into small segments so a player can change quality as network conditions change. The same edge architecture can also support edge computing, where short programs modify requests, authenticate users, or generate responses close to the requester rather than at a central data center.⁵

The apparent location of a CDN response is not always the physical location of the user or content. Routing choices reflect peering, congestion, capacity, policy, and failure conditions, and a nearby edge can still be slower than a farther but better-connected site. CDNs also raise questions of market concentration and network neutrality because large providers operate substantial private infrastructure alongside the public Internet. Their history is closely associated with the growth of multimedia-rich websites and the need to scale delivery beyond a single origin server.⁶

Origin server

The authoritative server or service from which a CDN retrieves content when an edge cache lacks a usable copy.

Edge server

A CDN server positioned near end users that handles requests and may cache or transform content.

Cache hit

A request satisfied by an object already stored at the selected CDN location.

Cache invalidation

The process of removing or marking cached objects stale before their normal expiration.

Point of presence

A physical or network location where a CDN operates routing, connectivity, and delivery equipment.