How does a content delivery network work?

---

When accessing a CDN, the own web server automatically connects to numerous other servers, which are called replica servers (exact memory images of other servers). The data of the own web server is then transferred to the replica servers via a distribution system. As soon as users access the website and call up certain content, a request routing system determines the replica server that is best suited to deliver the content. The location of the visitor is the most important factor here. For example, if the user is in the United States, the replica server should ideally be located in a state in the United States. The number and distribution of servers as well as their transmission capacity varies from provider to provider. Companies should therefore carefully consider which solution is most suitable for their own purposes. 



Source: https://www.youtube.com/watch?v=InhGahmVjdk

The two critical components in content delivery are the distribution system and the request routing system.

On the one hand, the distribution system helps to ensure that the original data is transferred to the replica servers. On the other hand, it ensures that the data is up-to-date and correct. The data is either sent directly to all replica servers or the respective replica servers simply receive a message that new data is available if required. However, the user requests still have to reach the correct servers, for which a request routing system is used. Visitors to a website do not notice whether the data originates from the source server or a replica server, as the entire process takes place in the background  

 

In order to find a suitable replica server for a request, the request routing system is guided by specific key figures. In addition to the best connection to the user, which is determined by the average transfer rate and geographical proximity, among other things, the utilization of the replica servers also plays a role. How the different metrics are ultimately weighted depends largely on the requested data. For example, if streamed media is involved, the priority is a stable connection. As soon as a suitable replica server has been found, the technical forwarding takes place, whereby several procedures are possible. The simplest is the client-side procedure, in which the user selects a replica server from a predefined list. An alternative to this is the server-side HTTP 302 redirection, which informs the user of a suitable Replica server that can be contacted automatically. However, the most common method is DNS-based request routing, in which requests are forwarded to a special DNS server within the CDN that provides the IP address of a Replica server.