In this article Wilco van Bragt describes the most known features and my opinion it's really needed within the infrastructure. he used the features available in VMware vSphere, Citrix XenServer and Microsoft Hyper-V.
High Availability (vSphere/XenServer) / Host Clustering (Hyper-V)
High Availability of Host Clustering is a feature that arranged when a hypervisor node (physical machine) fails the current Virtual Machines (VMs) are automatically restarted on one of the other nodes. Because normally a hypervisor node will host several VMs this is a very useful feature. Although the functionality is for a time frame not available for the end user, the system administrators don't have to start all the machines manually. Remember that you also need to have knowledge of Fail Over Clustering when using Hyper-V (this is the technique arranging this features).
vMotion (vSphere) / XenMotion (XenServer) / Live Migration (Hyper-V)
My personal favorite. With vMotin/XenMotion/Live Migration it is possible to transfer a running VM to another hypervisor used without (minimal) interruption from the end user perspective. Ideal for maintenance on the hypervisor host it selves or for manually arranging the load on the available hypervisors. I think this is the most used features within the server virtualization products.
vCenter (vSphere)/ XenCenter (XenServer)/ Virtual Machine Manager (Hyper-V)
Management of the hypervisor infrastructure is just like any other big infrastructure crucial. All products offer (a paid) central management suite, which is actual a requirement to maintain an infrastructure with more hypervisors a central management feature is a requirement. A management tools also offers additional features like role access and is for some other features a requirement.
Distributed Resources Scheduler (vSphere) / Workload Balancing (XenServer)
This feature continuously monitors the hypervisor infrastructure. If a host is getting overwhelmed this feature moves VM or VMs to the other hosts, so the resource usage is spread over all hosts. In theory this sounds perfectly, but in practice I see that this feature is doing nothing (because there is no high load) or the infrastructure is oversized and VMs are continuously moved between the servers (no practical experience with XenServer, I have seen this only on vSphere). However DRS can also be used to host VM's on a specific group of hosts (on vSphere, don't know if XenServer this supports) to arrange dividing for example VDI Desktops and Back-end Server, while arranging that when a host fails the other group of hypervisors can be used.
Remember that some other features are dependent of this feature.