Physical Machine Backup, Shallow Virtualization Backup, and Deep Virtualization Backup
ComputerWeekly.com’s Antony Adshead wrote a great article entitled “Top Five Issues in VM Backup” in which he called out the following top five issues with respect to virtual machine backup:
- Don’t treat virtual machines as physical machines.
- Make sure to manage application synchronization.
- Manage snapshots efficiently.
- Make sure to test restores.
- Track your virtual machines.
This is excellent advice. You should follow all of these tenants. What I found the most interesting in the article was the first rule: “Don’t treat virtual machines as physical machines.” This is true; however, there’s also a deeper issue that many of our customers have raised to us concerning the ability to perform granular backup at both the hypervisor level and within the virtual machine itself. These users are embracing a concept we call shallow virtualization versus deep virtualization.
Shallow virtualization was innovative several years ago as a mechanism that allowed virtual machines to be protected at the hypervisor (e.g., VMware vSphere, Microsoft Hyper-V, Citrix XenServer, etc.) level rather than within the virtual machine itself (also called the guest operating system level.) It was clearly a step forward in virtualization protection and made it much more convenient to backup virtual machines as “cattle rather than pets.”
The problem? As virtual machines have become larger, and as important applications have migrated to virtual machines, the practice of treating a large and very busy virtual machine the same way you treat all of your other virtual machines has become problematic. The most common complaint we hear in this area is when someone uses a shallow virtualization backup approach with a busy virtualized Exchange server and the virtual machine simply freezes for a few seconds — and the entire dependent infrastructure goes haywire and users begin complaining.
Deep virtualization backup solves this. In deep virtualization backup, if you have thousands of virtual machines, you select the few that have high change rates and you protect those virtual machines with software that understands the applications and operating systems within the virtual machine. This allows backup to occur at a more granular level without disrupting the environment or the users. A side effect is that it allows much better deduplication than any other technique if you have content-aware deduplication.
Have you seen any cases in which simplistic shallow virtualization only techniques have wreaked havoc?