VM Data Protection Primer: Comparing Data Protection Strategies for Virtual Server Environments

By Jay Kramer
published: Monday, January 18 2010

Server virtualization technology provides compelling business benefits by enabling companies to consolidate many physical servers into a smaller number of highly utilized servers or blade racks. In a virtual server environment, resources such as processing, application space, and storage are allocated by need, not by physical limitations. Now companies can buy and allocate computing resources in a way that increases operational flexibility, optimizes server performance and storage capacity and requires fewer administrators to manage.

However, without a well-planned backup strategy to protect a virtual server environment, the cost savings achieved can be quickly lost to increased complexity, increased risk and increased capacity requirements.

There are a variety of backup methods that can be used to protect a virtual server environment. This paper provides a framework for helping you define your data protection requirements to choose a backup strategy that best meets those requirements. 

In the world of physical servers, data protection is simple.  Each server is either a LAN client of a media server and is protected through either a LAN or SAN backup. In a virtual server (VM) environment, there is no longer a one-to-one relationship between the physical hardware and what is defined as a “server”. In addition, because VMs are easy and cost-effective to create, many virtualized data centers experience server proliferation. In these environments with many times more servers, the backup environment is significantly more complex.

Define Your Backup Requirements

Today, there is no industry standard for how to backup VM environments and there are significant trade offs to be made with every solution. Therefore, you need to start with a clear understanding of your backup requirements as well as your priorities (must-haves vs nice-to-haves). Consider the following your priority list:

• Backup performance: Start with an assessment of how well you are currently meeting your backup windows. Some methods of backing up VM environments can be I/O intensive (slowing performance) and some deduplication technologies can further slow down backups. Consider how much performance you are willing to give up for the benefits offered by your planned backup environment. Do you need to manage your I/O environment to avoid bottlenecking a given VM host?
• Restore performance: Do not assume that fast backup performance automatically means fast restores. Restoring data to a virtual server environment may be many times slower than backing it up. Determine whether you need single file recovery or will full volume restore suffice?
• Application consistency: Backup consistency may suffer when you are using a backup application with memory resident components. Consider how your backup software features and functionality will be affected.
• Added complexity: Many VM backup environments require the addition of proxy servers, intermediary backup software, and other new technologies. Weigh the cost and added complexity of these technologies against their relative value.
• Impact of backup volume growth: The larger your backup volume, the more complex and costly your backup environment is likely to be. Consider how quickly your data is growing and the impact of that growth on your backup environment. For example, can you store all of your data in a single backup system or will you be forced to divide it among multiple systems.

Understand the Tradeoffs

Client-based Backup
This backup methodology is the same as performed in the physical server world.  Each virtual machine is treated as backup client and will backup over the LAN as scheduled in the backup application.

Console Backup

In this model, the administrators use an application to protect the virtualized environment that is independent of the backup application, such as PureDisk or Avamar. These applications deduplicate the VM data and only send the unique new data to their master server. This model typically requires a separate management server and storage that is independent of the traditional backup environment.

Tradeoffs: Depending on the licensing model, this approach could save some licensing costs and it will minimize LAN bandwidth. It also provides capacity reduction through deduplication. However, you cannot use standard enterprise backup software with these solutions. As a result, application consistency may be difficult to maintain. It also creates a challenge around disaster recovery since console backup applications will typically need to replicate their deduplication repository for disaster recovery purposes.

VMware Consolidated Backup (VCB)

With the VCB model, you use VMware’s snapshot functionality to take an image of a running VM and the image is then presented either as a share or an entirely separate windows proxy server.  The proxy server is then backed up using a traditional backup application such as NetBackup or TSM.

Tradeoffs: This model is endorsed by VMware and it minimizes backup impact on physical servers. However, it can add considerable complexity in managing the snapshot and mounting processes. The challenge of managing the proxy host and VCB mounting process can be difficult. In addition, the snapshot functionality primarily supports Windows and can be difficult to implement in other environments. Note that the number of VMs per server is limited to 20 due to I/O constraints.

Storage Level Backup

Storage level backup uses storage arrays to snapshot existing volumes. Snapshots provide the benefit of instant creation and recoverability, but in order to maintain snap consistency VMware integration is required. It enables you to retain multiple snapshots online.

Tradeoffs: This is only supported by a limited number of array vendors and it can be costly to purchase snapshot capacity and licenses. You must be able to replicate snapshots for disaster recovery and retention may be limited by the number of snapshots supported by the array.

VMware vSphere

vSphere is a new VMware backup application with a complete list of changed data allowing you to peform a true incremental backup. It provides the fastest incremental VM backups and a simper recovery process.

Tradeoffs: Currently only supported by a limited number of backup applications.

Conclusion

There is no single approach to protecting virtual environments that is right for everyone.  Each approach has benefits and challenges and it is up to IT to decide the best option for their environment.  The first approach is the simplest and most straight forward and in practice, the vast majority of virtual server users are relying on this approach.  Most end users start with the first model and look to other options when they encounter problems or are planning on adding new hardware/software solutions to their environment.
 

SEPATON