From a machine/protocol perspective, how are paired-end reads treated differently from single-end reads by the Illumina Genome Analyzer?

My background isn't in biology, so I find the documentation a little difficult to follow.

In particular, I'm interested in the following:

• Are two separate clusters generated on the same flowcell for each fragment, one corresponding to each end?
• Are two sets of images output by the system, one for each end?
• Where does the Illumina Paired-End Module fit in? Is it used to generate the clusters differently, or does it have something to do with the imaging?

I am happy to split these into separate questions if people feel they cover too many topics, although I suspect an explanation might cover them all.