This page discusses proxy caching issues in more detail.

Figure 1: Web caching with a proxy server


Imagine two people at a company office -- let's call them John Doe and Jane Doe -- surfing the Net for business research. Suppose John has an interest in computer networking books and, having heard of the "In A Nutshell" series by O'Reilly, visits www.oreillynet.com in an attempt to learn more about them. It turns out that "O'Reilly Network" (the title of this site) does not cover O'Reilly's networking titles as much as it does some specific non-networking technologies. After fishing around a bit longer, John finds his ultimate destination at www.oreilly.com and merrily continues on his way.

Now it's Jane's turn. Jane is very interested in Python programming and hears that O'Reilly recently opened their online Python resource center. She navigates to www.oreillynet.com and, because this page was cached during John's very recent visit, she is surprised at how quickly this content-rich page pops into her browser window. With a great first impression, Jane is now ready to immerse herself in the wonderful world of Python development.

The potential benefits of proxy server caching loom even larger if John and Jane have a few hundred coworkers that share the same proxied Internet access and similar interests or Net surfing patterns. Yet proxy caching is not a silver bullet. Limitations exist that can render this technology much less useful.

Drawbacks of Proxy Caching

It's reasonable to expect that proxy servers handling hundreds or thousands of Web clients can become a network bottleneck. In addition to using servers with power processors and large amounts of memory, administrators may also choose to deploy multiple proxies to help avoid potential bottlenecks.

A proxy hierarchy creates multiple layers of caching support. Clients connect directly to a first-level caching, and if a Web page is unavailable there locally, the request "misses" and automatically gets passed to a second-level caching server, and so on.

As with many caching systems, the effectiveness of a multi-proxy server hierarchy is very dependent on the pattern of traffic. In the worst case, all clients will be visiting Web pages completely unrelated to each other, and proxies (the hardware, and the additional network traffic they generate) become pure overhead. One would expect that normal traffic patterns will usually not be worst-case, but every network's use pattern will be different.

Proxy caching differs from browser caching. Browsers automatically cache pages on the client computer, whereas proxies can also cache pages on a remote Web server. Because browsers already perform their own caching, introducing proxy caching into a network will have only a second-order effect.

Proxy caches don't help much with refreshed pages. On some sites, Web pages are set with HTML META tags to expire quickly; expired pages force the proxy cache to reload that page. Similarly, caching is rendered ineffective by pages that change content frequently, such as those on news sites, or weblogs.

Proxy caches also introduce measurement uncertainty into the Internet. Normally, a Web server log will record identifying information of visiting clients such as their IP addresses and domain names. For clients with proxy servers, all public requests are made on behalf of the server, using its IP address and identity. Web sites that carefully track the patterns of use of their visitors have much more difficulty in distinguishing unique client visits through proxies.

Next page > Proxy Servers and You > Page 1, 2, 3, 4, 5, 6