Caching 101: The Basics
(continued)

Q: Can't I address these problems by adding bandwidth?

A: To a degree, but two factors make caching a more effective alternative: First, storing popular content on a local disk (which is what caches provide) is much cheaper than buying extra bandwidth to repeatedly fetch that content. Secondly, adding bandwidth does not solve the latency problem experienced on the world wide 'wait'. Congestion on the origin server and other Internet infrastructure contribute to the latency of transporting objects over long distances, so a local cache server will improve access times.

Q: How do caches know how to handle an object?

A: All caches have a set of rules and all Web objects have a set of headers that are used to determine when to cache an object and when to serve it. The most common rules for caching objects are: 1. If the object's headers tell the cache not to keep the object, it won't - this is true for objects that are authenticated or secure. 2. If the object's headers and the cache's rules determine that an object in the cache is no longer fresh, the cache will check with the origin server to see if the object is still valid. Stale content will be refreshed - the updated objects will be fetched from the origin server.

Q: Where do ISPs typically deploy caches?

A: ISPs deploy caches at key nodes in their infrastructure, such as at the main data center and at the POPs.

Cache at POP brings data closer to end users.

Q: How do ISPs decide if they need a Web cache?

A: ISPs should look at their bandwidth costs and how frequently they have to add to this infrastructure to accommodate more end users, as well as the increase in the length of time their clients spend on the Web. These costs should be compared to the alternative cost of deploying caches to save bandwidth.

Most ISPs will see an ROI from a cache within the first few months of deployment. ISPs should also review their end user access times and determine the importance of investing in these improvements. In addition, caches are another element in an ISP's network that needs to be managed. ISPs should look at the deployability and ease of administration when considering a cache.