Backing up virtual machine data

Several suppliers of backup and/or disaster recovery software for virtual server environments have spoken with me. It seems that, like in any other are of computing, there are different approaches to backing up (and recovering) applications and data within a virtual machine and to deal with needs for disaster recovery. Here's a quick list of the approaches that seem popular.

• Stopping virtual machine execution, making a snapshot copy, restarting the virtual machine and then backing up the copy. Although this seems easy, it means that a production process has to be taken down for a time for the snapshot to be taken. It also means that in the case of a disaster, it would be necessary to rebuild the entire supporting environment before an individual virtual machine image could be restored.
• Running backup software within each virtual machine to save applications and data. As with the previous approach, it would be necessary to bring down production to back up applications and data. It also means, in the case of a disaster, it would be necessary for the IT administrators to rebuild the entire underlying environment prior to being able to restore the backup of applications and data.
• Use a product that can take snapshots of a running virtual machine and store only the changes that have occurred. It would then be possible to restore the current state by starting with the original image and updating it with all of the changes. It's clear that this approach also has issues if a quick recovery is necessary.
• Keep applications and data on a storage server and use the storage server's replication capabilities to keep backup copies or mirror images of applications and data. This approach, by the way, would also reduce the size of the virtual machine files.
• Use orchestration/automation software to keep a virtual server alive even though the underlying hardware was failing.
• Use a fault tolerant hardware configuration so no failure are seen by users. (This means totally redundant systems and storage infrastructure has been deployed)

I would suspect that large organizations are using combinations of these approaches to address their requirements for archival storage, uptime and infrastructure complexity.

What is your organization doing?