Computing Research: Are backups a brake on the virtualisation revolution?

The virtualisation revolution in the datacentre continues apace, as traditional drivers – consolidation of data, servers and licences and improved disaster recovery – are joined by a relative newcomer: the move to cloud computing.

A recent survey by Computing found that a half of datacentre servers remain unvirtualised, but since Gartner and others suggest that IT managers should plan for at least a doubling of virtualisation to meet demand for new servers within their organisations this is likely to change.

Another major growth area is application virtualisation. In April 2010 Computing found that only 21 per cent of firms had virtualised any mission-critical application. This figure is now 81 per cent; however the proportion of applications virtualised remains small.

We have also found an increase in the use of multiple hypervisors in order to optimise the performance of individual applications.

While there remains much scope for further virtualisation, to avoid exchanging infrastructure sprawl for virtual machine (VM) sprawl firms need to rationalise their backup and restore processes. Indeed some IT managers are already finding that backup issues are placing a physical limitation on their plans.

Backing up applications and data from an environment that includes multiple VMs and hypervisors can be a real challenge. While physical systems are often backed up using agents running inside the operating system, this may not be the most efficient approach for VMs. Network and disk I/O operations and server CPU loads are all increased while the backup is running, depriving other VMs running on the same host of resources. This can seriously diminish performance.

There are also network issues. Backing up VMs is a bandwidth-intensive operation that can lead to network bottlenecks.

With a quarter of respondents backing up more than 10TB of data every month, these performance issues cannot be ignored. Both the volume of data and the proportion residing in VMs are on the rise.

Databases were found to be the application most likely to be virtualised, making reliable backup even more crucial: databases often contain confidential customer data and other information covered by regulations such as the Data Protection Act (DPA).

So, what is to be done to ensure the continuing health of the revolution?

While it can be expensive, deduplication technology is often a good investment, especially if a large number duplicate files exist across the organisation.

Strain on the infrastructure can be reduced using serverless backups or SAN snapshots, and migrating live VMs to an alternative host prior to backing them up can also help reduce I/O overload.

It is also worth looking at virtualisation management products that manage both physical and virtual resources, or integrate with third-party backup software.

Then there is storage technology. Disk-to-disk-to-tape (D2D2T) processes offer a balance between short-term file recovery and long-term archiving, while running automated continuous data protection (CDP) tools in virtual environments.

No two virtual environments are the same, but adopting some combination of the above will keep the virtualisation revolution rolling.

Download the full report and analysis