Sites are the administrative container. Sites have SiteAdmins which are usernames in the Fedora Account System. Such people may edit the details of their Sites. Sites, via SiteToSite can also give read-only access to admins of other Sites, for purposes of seeing the site-private URLs among other things. Sites may be marked Private (e.g. for a company-internal mirror). As such, they won't appear on the public lists. Sites can be temporarily marked inactive, e.g. if you need to take a host down for maintenance.
Sites have one or more Hosts, which are machines serving content to end users. Hosts may also be marked Private. Hosts get their data by pulling from one of the master rsync servers. The master rsync servers require a DNS name or IP address to be in their Access Control List. As such, each Host should create one or more ACL IPs. Hosts can also be temporarily marked inactive.
Hosts carry content by Category. Fedora's categories include Fedora Linux (Fedora 7 and newer). Hosts expose a Category via one or more URLs (public URLs for anonymous http/ftp/rsync, or private URLs for use by other mirrors only).
A HTTP/FTP crawler will scan the public active sites every few hours to update its database of who has what. That crawler reports its User-Agent as mirrormanager-crawler.