How virtualisation affects backups

Server virtualisation presents us with a new problem in the backup arena. How do we capture information, and what strategy do we adopt? SMBs have three main options: Backup at the host level, backup within the virtual machine and backup via snapshot (VMware Consolidated Backup).

When running a virtualisation project, the majority of mistakes usually occur when a process, workstream or department hasn't been consulted in the initial planning and design phase of the project. This may be due to virtualisation being seen as a quick win for the server infrastructure team or as cost avoidance for desktop through virtualisation.

When it comes to backing up the virtual infrastructure, this can be the main area of pain, especially when constraints already exist. Virtualisation can then negatively compound the issue.

Server virtualisation enables us to squeeze the most out of our server/storage/network infrastructure. However, it presents us with a new problem in the backup arena. How do we go about capturing this information, and what strategy do we adopt?

There are typically three main options for most small- and medium-sized businesses (SMBs):

  1. Backup at the host level
  2. Backup within the virtual machine (VM)
  3. Backup via snapshot (VMware Consolidated Backup or VCB)

The first will put a backup agent on the ESX host and allow the host to back up its visible logical unit numbers (LUNs). The benefits are the simplicity of the rollout and the low level of additional hardware required to deploy this scenario. The challenge is that this places a high-resource requirement on the host itself, which can negatively impact the performance of the VMs. In addition, there's no support for database backups at this level, nor file restores, and the overall performance of the backup is low.

The second option is to place an agent in the virtual machine itself. This is ideal for organisations that wish to continue with the simple but familiar route to backup, and provides the granularity that has become expected of the process to date. However, this does require continued maintenance and management of each client, causing VM resources to be swamped during backup. This can also affect resource pools in the virtual infrastructure, with the bottleneck becoming network bandwidth to the media server.

The third option is to deploy a smart set of tools/services that support VCB or snapshots. These minimise the resource impact on the hosts and virtual machines, while increasing the restore options down to the file level and the entire VM instance. Management becomes simplified and processes are optimised. The disadvantages are increased requirements on shared storage (iSCSI/SAN) and an additional Proxy Server. For databases, pre- and post-scripts may be required for live (hot) backups.

However, when looking up the service maturity model, some enterprise organisations will have separated the data sets from the service (server/application) layer and applied an information lifecycle management (ILM) or content management and archiving system; therefore, the "information" or data layer is already protected, and the challenges move to service availability and compliance.

About the author: Andrew McCreath is the virtualisation practice lead at GlassHouse Technologies (UK) Ltd., a global provider of IT infrastructure services. McCreath has more than 16 years of experience in infrastructure and management information systems. Prior to joining GlassHouse, McCreath managed multimillion dollar projects while employed at Accenture, Credit Suisse First Boston, Kimberly-Clark, Société Générale and EMC. He currently specialises in server virtualisation and data centre consolidation.

Read more on SAN, NAS, solid state, RAID