Mounting a virtual machine image can save time, but there's a penalty to pay for the convenience. We list the questions to ask when deciding which data protection method to employ.
Backup software today is not the file backup software of yesteryear. It has evolved significantly to back up virtual machines (VMs); containers; cloud applications; edge devices such as laptops, tablets and smartphones; websites and more. But one of the data protection methods generating a lot of excitement is the ability of backup software products to mount a virtual or physical machine image directly on the backup or media server as a VM and put it into production. This capability is a game-changer when it comes to system-level recoveries because it enables server systems to be brought back in minutes -- a system no longer has to be recovered on new hardware and then put back into production.
So when does it make sense to mount a backup versus recover from a backup?
Challenges with mounting backups
Mounting backups and running them on media or backup servers decreases recovery time objectives (RTOs) by orders of magnitude, but the technology is subject to real limitations. For example, backups mounted on a media or backup server will often run in a degraded state, although exceptions exist when there is a separate physical host for those mounts. Media or backup servers are sized for backups, not application production and ongoing backups, so organizations should ask and answer the following questions while planning their data protection methods:
- How often do physical hosts fail?
- Will the budget allow for overprovisioning media or backup servers?
- If not, can applications, application owners or users tolerate running in a degraded mode if there is a physical host failure?
- How much degradation can users tolerate?
When answering these questions, many data protection professionals will envision a single VM running concurrently with the backups on that media or backup server. But that's an unrealistic scenario today. Should a physical host die, there will be far more than a single VM that will have to be brought up on media or backup servers. Additionally, most media or backup servers back up more than one hypervisor-equipped host plus other nonhypervisor-equipped hosts. Should there be a multiple host outage, each media or backup server will likely have to mount several dozen VMs concurrently.
Data protection methods: Choices for recovery
Organizations must also determine the types of recoveries most likely to occur in the data center. Anecdotal historical analysis of dozens of data centers, and especially backup and recovery service providers, has revealed that the vast majority of recoveries are a single file. There are two choices when recovering a single file:
- Go into the backup directory, find the point in time that the file needs to be recovered from and recover it. An example of this on a user level is Apple's Time Machine.
- An administrator mounts the point-in-time image of the machine with the file to be recovered, locates the file and recovers it. This option is a bit more burdensome because it requires the administrator to first mount the image before the file in question can be located. After recovery, the image must then be unmounted.
In general, both types of recoveries are required. It can be quite onerous to mount images every time a file needs to be recovered, but file recoveries cannot provide anywhere near the physical host failure RTOs that mounting images can offer.
The best data protection backup product can be mounted for physical host system and hypervisor-level recoveries -- assuming they have been properly sized and planned -- and perform general, file-level recoveries.