Friday, August 27, 2010

The initial challenge with protecting the VMware environment is the speed at which new servers can be deployed - from test, to initial deployment, to full scale production. The rapid adoption of virtualization is often underestimated. Mostly this is caused by the unexpected growth in the quantity of virtual machines (VMs) that often occurs as the project begins to prove itself viable. In these situations, the backup administrator is often left scrambling to get their arms around this new challenge. As a result, by far the most common method of protecting virtual servers is to back them up as if they were still physical, stand alone servers by installing a backup agent in the guest OS of the VM, not surprisingly, called “guest OS backup”. Of course these backup agents are no longer installed on physical stand alone servers, but are installed in virtual servers sharing a physical host. Because of virtualization there’s a high degree of data similarity between the virtual systems, making them ideal candidates for a deduplication system. While this similarity also exists in the pre-virtualized environment there are two important differences in the virtualized environment: server proliferation and host contention.

As mentioned earlier newly virtualized environments face rapid growth in virtual server population. Virtual servers can be added at almost no cost to the data center, thanks to plenty of excess physical host resources and the now-common practice of site licensing operating systems. There is nothing to order, nothing to rack and even much of the software installation is automated through the use of vVM templates. The impact is that there are more servers, (although virtual) than ever for the backup process to protect. The good news is that there is also greater justification for deduplication. Where deduplication may deliver a 20:1 storage efficiency rating in the typical home directory backup scenario, it can deliver 40:1 or more in a VMware environment because of this level of redundancy.

Having a deduplication storage system in place greatly eases a key pain that the backup administrator has to deal with: unexpected growth in servers consuming backup capacity, often in the form of tape. Second, there is already enough change going on in the environment, making the existing software better is a welcomed option. An application neutral deduplication system allows the backup administrator to stay with the existing backup software, add deduplication, meet expanded capacity requirements and survive this initial wave of the ‘server virtualization storm’.

Another advantage that application neutral deduplication brings to the virtual infrastructure is its ability to replicate data to a secondary site in case of a disaster independent of the applications used. While again, this capability is available even in stand alone systems, it’s enhanced in the virtual environment. Replication of these backup jobs is made more efficient because of all the redundant data. Products like Data Domain deduplication storage systems only need to send the unique data across the WAN even if that data comes from different volumes, different storage systems or even different sites.

At the remote site there is no longer a need to have dozens of severs to provide restore targets thanks to server virtualization. In the virtual environment a physical host or two can accept the restoration of several dozen VMs while additional physical hosts are being ordered. When those new hosts are available, features like VMotion and Storage VMotion can be leveraged to live-migrate those hosts to a new server or storage platform. As Storage Switzerland discusses in its recent article on direct recovery, products like Data Domain systems present storage as an NFS mount and are potentially capable of booting the VMs directly from the deduplication system.

As the environment evolves and the backup administrator has time to get their arms around the scope of the server virtualization, they can begin to look at the VMware specific extensions for their backup application or even VMware specific backup applications. Again the deduplication backup system is an ideal upfront investment since it can also support these new capabilities or products as they come online, since almost all of them back up data directly to a disk backup device.

A final point on the advantages of application neutral deduplication is that not everything is virtualized. In fact, several studies show that up to 70% of the servers in the typical data center are not virtualized. These stand alone, physical servers need to be protected as well. Since the application neutral deduplication appliances are agnostic to the source of the data being sent it can provide redundant data elimination across platforms, operating systems and even different backup applications - virtual and physical. This allows for the capacity savings, backup and recovery performance improvement and disaster preparedness functions to be universally beneficial. And an investment in technology to solve a backup problem allows the organization to both virtualize more servers and applications and to improve the backup process throughout the physical data center.

George Crump, Senior Analyst