FusionCompute Storage Virtualization

What it is

Storage virtualization in FusionCompute hides the type of physical storage behind a common object, the datastore. VMs get virtual disks from datastores and do not need to know whether the space comes from a SAN, a NAS or a local disk.

This lets the administrator create, grow, snapshot and move disks with the same steps on any storage.

How it works

Storage resource, storage device, datastore

FusionCompute uses three layers of storage object. Learn them in order.

For IP SAN, each host first needs a storage interface on the storage plane, so it can reach the array. The administrator then adds the storage resource, scans for its storage devices and creates datastores on them.

Virtualized and non-virtualized datastores

A SAN storage device can be used in two ways.

VIMS lets many hosts of a cluster read and write the same datastore at once. That shared access is what HA and live migration depend on.

Disk configuration modes

When a virtual disk is created, its configuration mode decides how space is allocated on the datastore.

Thin provisioning lets the total disk size given to VMs exceed the real datastore capacity. If real usage then fills the datastore, VMs can no longer write, so alarms on datastore usage matter.

Disk types and disk modes

A disk's type says who can use it. A common disk belongs to one VM; a shared disk can be attached to several VMs, for example for clustered databases.

A disk's mode decides how it behaves with snapshots.

Independent nonpersistent disks suit training rooms or kiosks, where every restart should return to a clean state.

Snapshots

A snapshot saves the state of a VM's disks, and optionally its memory, at a point in time. After a snapshot, new writes go to a separate area, so the VM can later be restored to the snapshot.

A snapshot is not a backup. It lives on the same datastore as the disk, so a storage failure loses both. Long chains of snapshots also slow disk I/O.

Storage migration and disk changes

• Storage migration moves a VM's disks to another datastore, while the VM runs or while it is stopped.
• Disk expansion grows a virtual disk; the guest OS must then extend its partition or file system.
• Disks can be attached to and detached from VMs as needs change.

A worked example

An administrator adds an IP SAN storage resource, scans it and finds LUN 3. She creates a virtualized datastore on it and associates it with every host of the cluster.

She creates a 100 GB thin provisioned disk for a new VM; the datastore shows only the space actually written. Before patching, she takes a snapshot of the VM. The patch breaks an application, so she restores the snapshot and the VM returns to its earlier state.

How the exam asks about this

• "Which disk configuration mode allocates space only as data is written?" -> Thin provisioning.
• "Which disk mode discards changes when the VM is shut down?" -> Independent nonpersistent.
• "What advantage does a virtualized datastore have over a non-virtualized one?" -> It supports features such as thin provisioning, snapshots and storage live migration.