Simply put, they are architectural terms that describe how Xen operates. Unlike the approach VMware takes, where the virtualization software offers an idealized hardware image to the guest operating systems as they are installed, Xen does not insert an idealized hardware image on top of the real hardware. Instead, it has a small software layer, which it calls the hypervisor, that mediates access to the real hardware beneath the hypervisor. You might think of the hypervisor as a traffic cop that directs hardware access traffic, coordinating all the requests from the various guest operating systems. In the past, guest operating systems that were designed to be installed directly onto the machine could not cooperate with the intervening Xen hypervisor; they needed to be modified so that hardware calls were redirected to the hypervisor function calls. This approach to virtualization was cleverly termed paravirtualization, although some people feel the term is overly-clever, as it causes confusion for people like you.
A new generation of chips from Intel and AMD have been released that offer the abilty to redirect hardware calls to a privileged software hypervisor. Intel calls this chip-based technology VT and AMD calls it SVT. This technology enables unmodified guest operating systems to make calls to the hardware which get redirected to the hypervisor, thus enabling the Xen hypervisor to coordinate multiple guest operating systems. In Xen parlance, this hardware-enabled virtuailzation is called HVM.
This was first published in December 2006