Squid (software) - Reverse Proxy

Reverse Proxy

The above setup—caching the contents of an unlimited number of webservers for a limited number of clients—is the classical one. Another setup is “reverse proxy” or “webserver acceleration” (using http_port 80 accel vhost). In this mode, the cache serves an unlimited number of clients for a limited number of—or just one—web servers.

As an example, if slow.example.com is a “real” web server, and www.example.com is the Squid cache server that “accelerates” it, the first time any page is requested from www.example.com, the cache server would get the actual page from slow.example.com, but later requests would get the stored copy directly from the accelerator (for a configurable period, after which the stored copy would be discarded). The end result, without any action by the clients, is less traffic to the source server, meaning less CPU and memory usage, and less need for bandwidth. This does, however, mean that the source server cannot accurately report on its traffic numbers without additional configuration, as all requests would seem to have come from the reverse proxy. A way to adapt the reporting on the source server is to use the X-Forwarded-For HTTP header reported by the reverse proxy, to get the real client's IP address.

It is possible for a single Squid server to serve both as a normal and a reverse proxy simultaneously. For example, a business might host its own website on a web server, with a Squid server acting as a reverse proxy between clients (customers accessing the website from outside the business) and the web server. The same Squid server could act as a classical web cache, caching HTTP requests from clients within the business (i.e. employees accessing the internet from their workstations), so accelerating web access and reducing bandwidth demands.