5) Understand the hypervisor model and performance consequences.
This is especially a big one for people that know Virtual PC and Virtual Server. The virtualization model that Hyper-V uses is very different from the model that Virtual PC and Virtual Server use. The Hyper-V model allows for much better I/O performance of virtual machines. This is mainly due to the new 64-bit hypervisor layer underneath everything - including the host operating system even, and the new high-speed VMBus "synthetic" drivers that run in the virtual machines.
Particularly- the use of optimized synthetic disk and network drivers talking to the VMBus, instead of using normal "hardware-oriented" drivers, make for a much faster I/O path from applications inside the Hyper-V virtual machines to the physical hardware. To make use of these synthetic drivers, make sure you use an operating system inside the virtual machines for which Microsoft provides so-called Integration Components. When you install the Integration Components, the synthetic drivers are installed as well.
Please see Microsoft Knowledge Base article 954958 for the list of operating systems in which you can install Integration Components.
4) Understand the use of snapshots
Snapshots in Hyper-V are very different from those found in Virtual PC and Virtual Server. Snapshots in Hyper-V allow you to save the current point-in-time state of your running or non-running virtual machine, and later come back to that particular state. Great for testing, troubleshooting and roll-back of virtual machine state.
This comes in the place of undo-disks, save-state, and to some extend even differencing disks with Virtual PC and Virtual Server.
Make sure you understand the power of snapshots and the scenarios where you should not roll-back the state of your virtual machine. Any scenario with a distributed database (such as domain controllers) is not a good candidate for snapshotting.
3) How to use Quick Migration
No discussion on virtualization can be complete without addressing fail-over support and virtual machine management. That is topic 3 and 2 in the list.
It doesn't take long to realize that any time you run multiple virtual machines on the same physical Hyper-V server, you have to think about how to handle the scenario where you have to do maintenance on the physical server (planned), or worse what happens when the physical server suddenly stops working due to loss of power or similar (unplanned).
For both the planned and unplanned scenarios, Hyper-V has support of host clustering. Windows Server 2008 clustering treats virtual machines as fully-managed clustered resources. For fail-over, clustering moves the
virtual machine from one node to another node. In Hyper-V terminology this is called Quick Migration. Due to the use of shared storage, only the content of the memory of the running virtual machine is copied to the other node.
2) Consider System Center Virtual Machine Manager
Virtual machines created with Hyper-V need to be managed as well. Enter System Center Virtual Machine Manager 2008. With Virtual Machine Manager you can simplify lots of tasks related to virtual machine management. This includes easier virtual host cluster support, automatically provisioning new virtual machines based on templates (including taking care of the "sysprep" part to make multiple virtual machines from the same template unique on the network), and an straightforward physical-to-virtual (P2V) conversion
process to move existing physical computers on to Hyper-V as virtual machines.
Interestingly enough, SCVMM 2008 can manage Hyper-V servers, Virtual Server, and even VMware ESX virtualized infrastructure, and even comes with a nifty virtual-to-virtual (V2V) option to move existing VMware virtual machines to Hyper-V.
As you would expect of a new server product from Microsoft, SCVMM 2008 fully supports automation with PowerShell.
1) Hyper-V can run on Server Core Installation
The number one thing to know about Hyper-V is the fact that it can run perfectly well on a Server Core Installation of Windows Server 2008. This means that on the physical server, you only need to install the absolute minimum "host OS,", and still have the full Hyper-V functionality. Having less moving parts and services running on the Hyper-V computer is, naturally, very beneficial to reduce the number of times you need to patch the server, and reduces the possible attack surface exposed to the network.
For any serious production installation of Hyper-V, and for any serious comparison with VMware ESX, being able to run Hyper-V on a Windows Server 2008 Core Installation is essential.