Check out the new USENIX Web site. next up previous
Next: Relaying front-end Up: Efficient Support for P-HTTP Previous: The problem with HTTP/1.1

Mechanisms for content-based request distribution

A front-end that performs content-based request distribution must establish a client HTTP connection before it can decide which back-end node should serve the request. Therefore, it needs a mechanism that allows it to have the chosen back-end node serve request(s) on the established client connection. In this section, we discuss such mechanisms.

The simplest mechanisms work by having the front-end ``redirect'' the client browser to the chosen back-end node, by sending an HTTP redirect response, or by returning a Java applet that contacts the appropriate back-end when executed in the browser [1].

These mechanisms work also for persistent connections, but they have serious drawbacks. The redirection introduces additional delay; the address of individual back-end nodes is exposed to clients, which increases security risks; and, simple or outdated browsers may not support redirection. For these reasons, we only consider client-transparent mechanisms in the remainder of this paper.


Peter Druschel