Best Practices for Integrity Virtual Machines

11/19

Storage Management and Fault Tolerance

Arguably one of the most powerful technologies in Integrity VM is its capability to use multiple entities

– files, logical volumes, disks, etc. – as virtual mass storage (virtual hard drives) for VMs. As

mentioned earlier, standard HP-UX tools can be used to manage storage on the VM Host. The same

is true for fault tolerant strategies and solutions. RAID strategies, such as mirroring, should be done

on the host. There are several reasons for this, including:

• Protecting the physical storage on the host automatically protects it for the VMs using that storage.

• Storage fault tolerance solutions need only be implemented once on the Host as opposed to

implementing them multiple times – once for each VM.

• Storage fault tolerance on the VM will often be a waste of time because it protects against

hardware failures which can not occur in virtual devices but only in physical devices.

To illustrate these points, consider the simple VM configuration in Figure 5. In this configuration, two

physical disks on the VM Host are partitioned into four logical volumes each, for a total of eight

logical volumes. Each of these is then used as virtual hard disks used by VMs – four VMs, each using

two logical volumes. On each VM, the two virtual hard disks are mirrored for data protection. The

problem with this approach is that if one of the physical disks on the VM Host should actually fail,

then the mirroring on two of the VMs provided no benefit whatsoever – all four logical volumes would

fail along with that physical disk.

Figure 5 – Data protection in a VM usually does not provide true fault tolerance for

storage – this is not a recommended configuration

Integrity VM Host

MIRROR

DISK 0

DISK 1

Integrity VM Host

MIRROR

DISK 0DISK 0

DISK 1

Contrast this scenario with that in

Figure 6 where data protection is implemented on the VM Host. With this approach, the physical

storage is protected and it need only be done once – on the VM Host. Similarly, mass storage arrays

(with their own data protection implementations) connected to the VM Host should be used in this

same way for the same reasons.