A content delivery network (CDN) is a network of computers that contains copies of media or other files, stored at different points of presence (POPs) on a network, to provide the maximum throughput or bandwidth to clients using the network. When someone visits your site, the CDN automatically decides the shortest and ultimately the fastest route to get that content to your visitor. Clients can get access to one of the copies of the data that is stored nearest them, as opposed to every client using date from one central server. This avoids bottlenecks and can dramatically improve performance.

Content can include media files such as movies (website video), software and documents, applications, or live media streams.

The capacity of multiple POPs can be more than its backbone capacity. Because of this, a significant increase in the amount of simultaneous users is possible. As an example, with a 10 Gbit per sec backbone and 100 Gbit per sec per server capacity, 10 Gbit per sec is the limit that can be served. With a CDN, though, 10 servers at 10 geographically dispersed locations (on the “edge”), the capacity can be significantly increased to 10×10 Gbit per sec.

Geographically dispersed POPs decrease the load on the standard infrastructure, allowing for increased capacity and lower costs.

Supposedly, geographic proximity to the content results in faster delivery, but distance isn’t necessarily the factor for best performance. Users at the receiving end viewing media such as website video are likely to experience less stuttering and improved stream quality, even more noticeably so in remote areas. This increased quality can provide HD quality content with a higher quality and a lower network load.

CDNs often have automatic load balancing with automatic redirection to a less loaded POP. Importantly, because of their inherent redundancies, CDNs can provide persistent availability even if network, power, hardware or other outages happen.

CDN POPs are usually placed in multiple locations, often over different backbones. These POPs work together to serve content requested by end users, in the background without apparent notice to end users, optimizing the delivery and user experience. End-user performance experience is consistently more optimal the availability of content overall is noticeably improved.

When there is a request over a CDN for media, it is directed to an optimal POP. This may be measured by choosing a location that takes the fewest hops, the least amount of “network seconds,” or the best availability.

Content Delivery Networks use web caching, server-load balancing, request routing, and content services.

Web caches contain commonly used content. This lowers bandwidth needs, reduces server overhead, and lowers the response times that end users experience for content being accessed from the cache.

Load balancing is done in one or more ways, such as global load balancing (hardware based), also known as a “switch” – as in web, content, or multiplayer switch. Traffic arriving at the switch is sent to a real web servers that is attached to the hardware switch. This balances load, increases capacity, improves scalability, and provides increased reliability. A failed web server can have its load redistributed, simultaneously allowing for an automated server performance check.