Data protection is a relentless problem for Windows systems administrators. The demands for greater uptime and faster restoration have often marginalized traditional backup methods in favor of point-in-time (PIT) snapshots. There's little question that snapshots bring real benefits like speed and simplicity -- two attributes that are notoriously absent from traditional backup approaches.
In spite of the benefits, however, snapshots are not a perfect solution to every data protection problem. Let's examine the importance of snapshots in today's user community and consider some pitfalls that administrators need to avoid.
Snapshot speed and simplicity
The importance of snapshots has been driven by two principal factors: time and virtualization. In today's 24/7/365 world, many organizations simply don't have the luxury of taking a server down for several hours to perform a traditional backup. In addition, the recovery point objectives (RPOs) for most mission-critical servers are decreasing to the point where even just a few minutes of lost data can be catastrophic to a corporation.
Todd Ericksonpresident, Technology Navigator
Server virtualization exacerbates this problem, as now a physical server can host numerous virtual servers. Quiescing even one of those virtual servers for an I/O-intensive backup cycle can seriously impair the performance of every remaining virtual machine (VM); backing up every VM simultaneously is simply out of the question.
Snapshots overcome these practical problems by creating copies of VM files directly from the server's memory space storage, or duplicating an existing VM file from SAN storage to another storage location (such as a disaster recovery site). Snapshots are generally quite fast, and full copies can usually be accomplished in less than a minute. An incremental snapshot can be taken in just a few seconds. This makes it possible to achieve very small RPOs while still protecting numerous VMs on the same physical host server.
Today, the popularity of snapshot technology is clear. In a recent virtualization survey conducted by TechTarget, over 63% of IT professionals reported using snapshot technology in their virtual environments.
The speed and convenience of snapshots have also extended their usefulness far beyond common backup and recovery roles to include test and development support. Creating a snapshot before an upgrade or patch allows an administrator to test the changes on real data without affecting a production server, or roll back the server if the change precipitates a problem. "The biggest benefit that we're finding is in change management, loading and testing things where you can always revert back to the original version," said Todd Erickson, president of Technology Navigator, a financial business intelligence firm in Cary, NC. "To me, that's been the greatest value."
Windows administrators can typically manage snapshots using tools native to the virtualization platform (such as Hyper-V Manager or VMware Data Recovery) or from a third-party. In our recent virtualization survey, respondents reported a variety of popular tools for Hyper-V including Citrix Essentials for Hyper-V, Symantec Veritas NetBackup PureDisk, Acronis True Image, CommVault Simpana Software, PlateSpin Ltd.’s Forge and EMC Corp.’s NetWorker, among others. In addition, the storage sub-system itself (such as FalconStor's VTL) may provide storage-to-storage snapshot capabilities.
Snapshot management is particularly important from the standpoint of automation. Users can certainly create, restore or delete snapshots on-demand for one-off or as-needed snapshots, but most Windows systems administrators want to hand over as many mundane functions as possible to the management tools. "We're just using the tools that come from the vendor," Erickson said. "They're allowing us to automate snapshot creation and removal and provide naming."
Even though 80% of virtualization survey respondents report using some form of SAN, experts point out the importance of tracking and managing the storage space allocated to snapshots. Snapshots accumulate, but their value declines with time, so limiting the space used by snapshots and deciding when to delete or overwrite old snapshots are critical management choices. Erickson suggests providing anywhere from 10% to 20% of committed storage space for snapshots. For example, if VMs require 2 TB of storage, set aside 200 GB to 400 GB for snapshots. It is possible to conserve storage space by reducing snapshot frequency, deleting old snapshots sooner, and using incremental snapshots in concert with full snapshots. Scripting and automation can simplify many of those mundane processes, but snapshot management for critical VMs may still be performed manually.
Snapshot integrity and backups
The main problem with snapshots is the concern over data integrity. In order for a snapshot to be valuable, it has to be recoverable, and this means it needs to have all of the data -- the precise state of the virtual machine -- at the moment the snap is taken. Unfortunately, this doesn't always occur with every Windows application. "Reverting back to the snapshot on any box that contains real-time information means that there could be a loss of data," Erickson said. "It can get really dicey when you start talking about anything that deals with Active Directory or Windows DFS or any of the Microsoft Active Directory services."
This means administrators will need to perform due diligence and test the snapshot/restoration process before trusting snapshots in a production environment. In some cases, an application may need to be formally quiesced in order to flush any buffers and ensure that the VM is in a steady state before executing the snapshot. In other cases, a troublesome application may demand an alternative data protection scheme.
"Does the snapshot include the particular data that you're looking for at a given moment?" said Chris Steffen, principal technical architect at Kroll Factual Data in Loveland, CO. "Until you actually deep-dive into the snapshot itself, you don’t know if that's the case." Ultimately, successful restoration will determine just how well snapshots support each particular application.
Over 35% of virtualization survey respondents cited backing up virtual machine data as the second greatest concern in a Hyper-V environment, just behind storage management. Snapshot integrity is perhaps the most pressing issue here, but snapshots themselves are also valuable data that may need to be protected with backups. For example, saving a VM state to disk doesn't protect the snapshot against a problem in the storage subsystem, so storage-based or LUN-level backups may also be needed to protect snapshots that have value over the long term. Backups can reside on another storage system locally or at another physical site connected across a WAN.
Stephen J. Bigelow, senior features writer, has more than 15 years of technical writing experience in the PC/technology industry. He holds a BSEE, CompTIA A+, Network+, Security+ and Server+ certifications and has written hundreds of articles and more than 15 feature books on computer troubleshooting. Contact him at firstname.lastname@example.org.