What is virtual machine backup?
Virtual machine backup is the protection of virtual machines by copying their virtual disks and configuration as a unit. It is usually done at the hypervisor level, without software installed inside each machine, and produces a restore point from which the whole machine or individual files can be recovered.
How does virtual machine backup work?
The backup software connects to the virtualization platform, requests a snapshot of the machine and reads its virtual disks through a backup proxy. Changed block tracking tells it which blocks have changed since the last run, so later backups copy only those. For machines running databases, application-aware processing quiesces the application first. When the copy is complete, the snapshot is released. The result is an image-level backup.
How does it compare with agent-based backup?
Both approaches are in use, often side by side.
| Aspect | Agentless, through the hypervisor | Agent inside the machine |
|---|---|---|
| Software to maintain | None per machine | An agent on each machine |
| New machines | Picked up by policy, tag or folder | Agent must be installed |
| What is captured | Whole machine image | Image or selected files |
| Works for physical servers | No | Yes |
| Dependency | Hypervisor snapshot interface | Guest operating system support |
What restore options does it provide?
An image of a whole machine supports a range of restores. The full machine can be restored to its original host or to another, including a different hypervisor with some products. Individual virtual disks can be reattached. Files and application items can be extracted through file-level and item-level restore. And instant recovery starts the machine directly from the backup repository within minutes, while its data is moved back to production storage in the background.
What does it ask of backup storage?
Routine backups write large sequential streams and are limited by network and repository ingest. Recovery is more demanding. Instant recovery and file browsing read randomly from the repository, and a mass restore after an incident reads hundreds of machines at once. Repositories that are quick to write to and slow to read from make backups look healthy and recoveries slow. Restore throughput, measured under parallel load, is the figure to check, and mass VM recovery planning covers how to order a large restore.
Frequently asked questions
Is a VM snapshot a backup?
No. A snapshot lives on the same storage as the machine and depends on it. A backup is a separate copy in a repository.
Do virtual machines still need agents?
Not for image backup. Agents are used where finer control inside the guest is needed or where the hypervisor is not supported.
How often should virtual machines be backed up?
As often as the acceptable data loss for each system requires. Daily is common, with more frequent runs for critical workloads.