This blog has been re-posted and edited with permission from Matt Glaman's blog.
Phil Karlton is quoted as having said, "There are only two hard things in Computer Science: cache invalidation and naming things." As someone who suffers horribly from the latter, I'm glad Drupal's caching APIs make the former a breeze. The long short of it is that caching of data improves performance by not having to perform the same operations multiple times after it has been done the first time until you need to re-run that operation again. It involves some storage (static variable in memory, memory backend storage, or the database.) Generally, you store the data with a given identifier for later retrieval, the actual data to be cached for later reuse, and possibly an expiration time for automatic invalidation.