Glossary

Guest
Guest is the term assigned to a typical virtual machine. It is a complete virtualized instance of Linux but it is limited in the hardware resources that it can access (limited by the virtualization software/host in use).
Host
Host is the term assigned to the management system for a virtualized server. Typically this is a specially priviledged kernel that can access raw hardware devices and can create/destroy virtual machines (guests). Common examples of hosts are UML host kernel, VMWare ESX Service Console, Xen Domain0.
HVM
Hardware-assisted virtual machine refers to VMs that are taking advantage of Intel-VT or AMD Pacifica extensions. May be a Xen specific acronym.
Hypervisor
Virtualization architecture consisting of a thin software layer that runs directly on the hardware intercepting some, or all, operating system calls to hardware. Typically the hypervisor virtualizes CPU and memory while a single priviledged guest is assigned to manage and virtualize network and disk.
NUMA
Non-Uniform Memory Access systems are essentially SMP computers with multiple memory busses. As the acronym suggests, access to some regions of memory is quicker than to others. Most often each CPU will have a range of memory that is considered local, the operating system must optimize to keep memory for processes close to the CPU that is executing said program.
OSS
Open Source Software refers to software that is made available in source code form and is usually free of charge. Not only is source code available but you are free to modify it for your own purposes and (in the case of some licenses) you may even resell OSS without any fees or royalties being owed to software’s creator. Examples of OSS licenses are MPL, GPL and BSD. Refer to the Open Source Institute for a better definition.
Pacifica
AMD Pacifica is supported in some CPUs available since 2006. Pacifica is a technology that builds some virtualization awareness into the CPU enabling the software virtualization layer to be less complex and higher performance. Intel has a similar technology named VT-x.
Paravirtualization
A virtualization technique where the operating system is modified to be aware of the fact it is virtualized. This has a number of benefits—principally a less complex virtualization layer and more opportunity for optimization since OS is aware of it’s envifronment.
PAE
32bit computers can only directly utilize 32bits of memory. This translates into about 4GB. While 4GB is sufficient for many operating system uses when virtualizing a server being limited to 4GB of RAM will severely limit the number of virtual machines that can be deployed. PAE support enables 32bit servers to indirectly access 36bits of memory (server motherboards typically support up to 64GB). 32bit servers with PAE support can often be configured with 32GB of RAM enabling a much greater number of virtual machines per server. Not all virtualization solutions support PAE.
SMP
Symmetric MultiProccessing computer systems have multiple CPUs that share a common operating system, I/O subsystem and memory bus. Large SMP systems offer very high performance but access to memory can be a bottleneck. NUMA is an attempt to address the problem of memory bottlenecks in large systems.
VT-x
Intel VT (Virtual Technology) is supported in some CPUs available since late 2005. This technology builds some virtualization awareness into the CPU enabling the software virtualization layer to be less complex and higher performance. AMD has a similar technology named Pacifica. For more details including a list of processors with VT capability see the XenSource Wiki.