Backup is a big job. Historically, it's meant moving much of an organization's data set from primary storage devices (servers or shared arrays) to a dedicated backup storage area. A lot of technology, such as snapshots and deduplication, has been developed to help address these issues, but for the majority of organizations, it still involves a significant infrastructure and an increasing level of complexity and cost. In addition, server virtualization has made backup even more difficult, prompting companies to look at virtual machine-specific backup software products.
Virtualization started a revolution in the data center with server hardware consolidation and the ability to migrate entire server instances from one location to another, even while applications were running. IT became used to creating new server instances whenever they were needed (provisioning), being able to clone common virtual machines (VMs) to speed that process and then move them to balance the load in the environment.
This move to virtualization certainly brought benefits to the IT organization, but not for backup. Before being optimized for virtual environments, existing backup applications had to treat each VM as a separate backup client, typically with its own agent and license for the backup server. While server virtualization was simplifying the IT environment, backup was becoming more complex and expensive.
It's not hard to see why IT organizations were eager for an alternative to back up their growing virtual environments, and several software products emerged to address this new challenge. Initially, they could only do image-level backups of VMs by writing code that ran in the VMware hypervisor's "service console," but when VMware released an API set to facilitate VM backup and restore, things became a lot more interesting.
VMware's vSphere Storage APIs for Data Protection (VADP) enables the backup application to conduct centralized VM backups without running an agent within each VM. The API enables the backup software to take a snapshot of the VM and use it to create image-level backups and restores, and to support file-level restores. With VADP and developments such as Changed Block Tracking (CBT), which enables the backup application to identify net-new data at the block level, these VM-specific backup products have some real advantages over legacy backup applications.
Why look at VM-specific backup?
One of the toughest jobs for an enterprise backup application is platform support. Every server or computer that is backed up needs its own agent, each of which has to be developed and tested with every version of every OS. Support for programs such as Exchange, Microsoft SQL Server and Oracle just adds complexity.
Legacy backup products eventually incorporated VADP and CBT, but were (and often still are) slower to match the functionality provided by these new VM-specific solutions. This was because these products' older, more complex architectures took longer to integrate the new API commands. Virtual machine-specific products also have the benefit of focus, with only one task to perform well rather than being a Jack of all trades. This makes it easier to bring out new functionality and respond more quickly to updates, since they're only dealing with a few hypervisor platforms.
The move toward VM-specific backup solutions can also be attributed to a shift in corporate IT organizations. As virtualization grew in popularity, the VMware administrator's role became more prominent. Similar to database administrators, these admins wanted more control over their virtual server domains, so to speak, which included data protection. When the question was raised as to how to improve virtual server backup, the decision to try a VM-specific solution was easy for them to support.
VM-specific backup product functionality
To begin with, these solutions simplify the job of backing up VMs and usually cost less. Since they were designed from the ground up to back up VMs, they're easier to implement and run, and don't require as many software licenses. And because they don't have to support the physical server environment and don't require all the legacy code of traditional backup applications, they're lighter and need fewer server resources.
Server virtualization enables some advanced functionality that products designed to back up VMs can leverage. These products can deliver nearly instant server recovery, provide application failover without clustering software, and support cloud-based backup and disaster recovery as a service (DRaaS). It should be noted that legacy backup solutions can include many of these features but, as mentioned above, the VM-specific products are typically first to bring these to market.
Recovery in place
While server virtualization has complicated backup to some extent, it has also brought a significant new benefit: recovery in place. Virtualization enables users to move workloads around the environment to balance the load on storage and compute resources. This portability also makes a server easier to recover.
When a physical server went down, using traditional backups to affect a recovery was a multi-step process. The server had to be "rebuilt" (typically reinstalling the OS and current applications) and then the backup data had to be restored to that server before it could be restarted. A bare-metal restore would speed up that process, but it still required moving all the system and data files back to the server.
With recovery in place, a hypervisor only needs to be pointed to the backup copy of a virtual machine (VM) image and it can restart that server instance for a nearly instant recovery. While running a production VM on the backup infrastructure will be slower than on production storage, at least the applications are back online.
For remote recovery or cloud-based disaster recovery as a service, the migration process back to the production infrastructure is more complicated and should be well understood before the product is put into production. While an application is running remotely, it's accumulating new data. This includes the time it takes to copy (or ship) that data set back from the remote location. All this data must be saved and then merged into the primary production system, which can be a sophisticated process.
As mentioned earlier, most backup solutions developed for physical servers now have modules to support VM backup and restore using VADP. Most provide the core capabilities of VM-level backup and file-level restore using snapshots, CBT and the ability to support recovery in place. But products designed exclusively to support VM backups offer a number of additional features and are becoming more popular, even for companies that already use a product for physical server backup. We'll now look at four of these products.
Veeam Backup & Replication
Veeam's software-only solution runs on a Windows server or VM, and supports VMware and Hyper-V to provide fast VM image-level backups using snapshots and CBT. Veeam also provides image- and file-level restores. The company claims the product can recover a failed VM in two minutes.
Veeam Backup & Replication leverages the same snapshots to create a local copy for backup and to replicate to another target for DR purposes. The snapshots support backups to tape and can provide file-level recovery from any virtualized application on any OS.
Veeam supports recovery in place and automatic backup verification, emphasizing the simplicity of this process and its ability to support a test and development space. WAN acceleration is included to reduce bandwidth for offsite backup copies, as well as embedded deduplication and compression to minimize storage consumption.
This is a software solution that backs up VMs from VMware only, running as a VM or as part of a turnkey backup appliance from Quantum. vmPRO backs up data in its native, non-proprietary file format, enabling users to browse the backup data store and see the VMDK files, plus the individual files resident in the VMDK.
This enables users to boot directly from the data store for recovery in place or to restore files without a backup application. vmPRO also supports the Linear Tape File System, the Storage Networking Industry Association standard LTO file system that enables drag-and-drop access to files recorded on tape.
vmPRO also includes a data reduction process called Progressive Optimization that goes beyond CBT. By leveraging some file system-awareness, Progressive Optimization can reduce the data footprint to improve data handling efficiency and speed. Quantum claims that, when implemented with Quantum's DXi deduplicating backup storage systems, vmPRO can offer up to a 95% reduction in backup data storage.
Unitrends Virtual Backup
Unitrends Virtual Backup (formerly PHD Virtual Backup) can back up VMs from VMware, Hyper-V and Citrix XenServer, installing as a VM on any of these hypervisors. Leveraging VADP to do snapshot-based backups and restores, Unitrends emphasizes simplicity; it claims only four clicks are required to complete a backup and another four clicks to recover.
The product offers recovery in place, but takes it a step further with a module that runs on a separate appliance creating "certified recovery points" for specific applications. This lets users test and confirm that backups for a single VM or for multiple VMs in an n-tier configuration are indeed recoverable.
Unitrends stores backup data in blocks, not in specific jobs, which allows the product to apply deduplication globally. This architecture lets blocks from older backups be included in the comparison process for each subsequent backup, generating a greater level of data reduction and lower storage costs over time. Unitrends also offers encryption for data at rest.
vRanger backs up VMs from VMware and Hyper-V, installing on a Windows server or as a VM. The product also backs up Windows physical servers by performing a physical-to-virtual conversion. vRanger supports source-side and target-side deduplication, integrating with Dell DR appliances and EMC Data Domain, with support for the DD Boost API.
vRanger offers full, incremental and differential backups with efficient remote replication for DR purposes. VMware Hot-Add for improved performance is supported, plus Active Block Mapping and CBT for improved efficiency and reduced storage consumption. It also offers Instant File-level Recovery for Windows and Linux, and provides encryption during backup and for data at rest.
Server virtualization is certainly here to stay and so, it would seem, are backup products designed specifically to handle VMs. They're generally faster, easier to use, more flexible and have a richer feature set than legacy backup products. They can also be less costly, since they don't require as many licenses and need less compute resources and storage capacity.
For companies that already have a traditional backup system in place, buying a virtualization module may be the best way to back up a growing VM inventory. That said, many organizations are still buying separate VM-specific backup solutions for the reasons mentioned above. For those without an existing backup system in place, a dedicated VM backup solution makes even more sense because it doesn't require the backup infrastructure that legacy products do.
About the author:
Eric Slack is an analyst at Storage Switzerland, an IT analyst firm focused on storage and virtualization.