Get answers about storage, network requirements, live migration, and other details about CSVs
1. Do CSVs require SAN storage or can they be implemented using DAS?—CSVs must be implemented using shared storage; they can't be implemented on DAS. CSVs can be created by using any of the shared storage technologies that are supported by Server 2008 R2: iSCSI, Fibre Channel, and Serial Attached SCSI (SAS). However, CSVs do require NTFS as the file system.
2. Do CSVs require Windows Failover Clustering?—Yes. CSVs are created by using Windows Failover Cluster Manager, and they must be created on shared storage that's visible to the failover cluster. Therefore, Windows Failover Clustering is a requirement for using CSVs.
3. Do CSVs have any special networking requirements?—No. CSVs and live migration work over standard Ethernet network connections. Microsoft recommends that VMs have one network connection for client systems that are accessing the resources on the VMs. Microsoft also recommends a dedicated network for CSVs and other live migration traffic. In addition, if you're using iSCSI storage, Microsoft recommends an additional dedicated network for iSCSI traffic.
4. When you perform a live migration, does the VM get moved to a different CSV?—No. Multiple cluster nodes can access a CSV simultaneously, so there's no need to move or mount the VM's VHD files when live migration moves the VM to a new host node. CSVs let the new host node immediately access the existing VHD files.
5. Are CSVs required for live migration?—No. CSVs and live migration were both introduced with Server 2008 R2 and virtually every example showing live migration uses CSVs, so it's easy to jump to that conclusion. To use live migration without CSVs, you need a separate LUN for each VM that will take part in the move. However, CSVs make live migration faster because there's no need to mount the new storage.
6. Are you limited to a single CSV per host?—No. By default, Windows Failover Clustering creates a single CSV mount point of C:\ClusterStorage\Volume1, but you're not restricted to using just that mount point, nor are you restricted to using that naming convention. You can create multiple CSVs and each must have a unique name.
7. How do CSVs appear to the host OS?—To the physical host system, the CSV is seen as a mount point in the standard file system. When you create a new CSV, the Failover Cluster Manager creates a folder for the CSV, which, by default, is found in the host's file system as C:\ClusterStorage\Volume1. If you create additional CSVs, they'll have different host folder names.
8. If you create a CSV, do you have to use it for all your Hyper-V VMs?—No. Using CSVs for your VMs is strictly optional. When you create a new VM by using the Hyper-V Manager or Virtual Machine Manager 2008, you have the option of storing its files on a CSV or on any other location that's accessible to the host.
9. How do CSVs appear to the guest OS?—To the VM guest, the CSV storage appears as standard storage. In other words, if the guest is created on a CSV and that guest is configured to use a single drive, the CSV storage simply appears to the guest as a C drive.
10. Are there any limitations to using CSVs?—Yes. CSVs are essentially like other host storage. However, CSV's don't support pass-through disks, which could be a significant limitation for some I/O intensive applications where pass-through disks can provide slight advantages in raw I/O performance. In addition, backing up VMs with Windows Server Backup from the host requires restoration of the entire image. Other backup solutions, such as Microsoft System Center Data Protection Manager (DPM) 2010, provide more options.