The term 'sprawl' seems to be gaining momentum in IT circles. The culprit, we are told, is virtualisation: while creating, duplicating and even cloning new virtual machines may be much simpler than procuring their physical equivalents, organisations can end up with so many of the things that they start causing management problems. IT isn't straightforward to manage at the best of times, and the usual issues of operations, patching and licensing can quickly get out of control if the number of systems gets out of hand.
One area in particular that can suffer problems is backups, and it's not rocket science to work out why. Backups require transfers of sometimes-large quantities of data, from system, application and user repositories: more machines equals more data to be transferred, which can increase pressure on physical server and network interfaces particularly if everything sis being backed up at once. Lose control of the systems in a sprawl situation, and you lose control of what needs to be backed up.
As a storage management software company, we are of course very happy to provide technologies to help organisations with their backups - and we continue to develop and enhance our products and services to meet your evolving needs, bringing in support for scenarios such as dynamic IT environments that make heavy use of virtualisation technologies. The V-Ray features we have built into Symantec NetBackup and Backup Exec are a case in point.
To address the challenges, technology needs to work with people and process. While these are early days for best practices to be fully established, here - in no particular order - are some emerging 'golden rules' based on conversations I've been having with customers. Feedback welcome, as always.
- First and as suggested above, timing is everything. If you have ten virtual machines running on the same physical box, it's probably not a good idea to attempt backing them all up at the same time.
- Second, one of the most straightforward ways of reducing the 'backup window' (the time taken to back up necessary data), is by avoiding creating data unnecessarily. In other words, not creating virtual machines. This requires virtual server provisioning processes to be in place, which can be no bad thing.
- Third, not all virtual machines are created equal. You can think about how you classify virtual machines - for example into critical/live/staging/development and set their backup criteria accordingly.
- Fourth, spread the load - keep on top of VM's not just from a processing standpoint but also the amount of data they require to transfer at backup time, and build this into the backup configuration.
- Fifth, look at the whole architecture. Virtual machines, and/or their data may exist on the physical machines or may be accessed directly from a direct- or network-attached storage device. Wherever physical data is situated, the principle of closing the gap between storage and backup still stands.
A final point is more about common sense than best practice - start now. While VM sprawl may not yet be causing you problems, you don't want it to become a problem that could later turn into a crisis. Avoiding virtual machine sprawl now may well be key to avoiding backup sprawl in the future, so the sooner organisations can start to implement good practices, the better.