It’s been an interesting time for services. Microsoft/Danger/T-Mobile lost some of their user’s data recently when an outage turned into a data loss. Users who hadn’t backed up their Sidekick data were out of luck. Many users assumed that because it was a cloud service, it didn’t need a backup.
I think this event highlights something that’s important to think about. Backups. The fundamentals don’t change in virtualization or in cloud. You had better backup your data or you’re going to be out of luck, regardless of the platform.
I’m pretty paranoid about backups, I’ll admit. Each of my desktops gets backed up via a combination of Home Server or Time Machine, and my handheld gets backed up to the desktop and that gets copied elsewhere. I use a “cloud” disk, which I take a backup of every day. All my servers have backups included as part of the service, and I get insight into those backups. We offer similar services to our customers for their systems. This level of paranoia comes from being burned. You lose data only once in your career before you learn.
Virtualization actually can add complexity to backups. I view backups at two levels, Operational and System. Your operational backups are used for pulling data that is lost. These are ideal for simple recoveries, such as a lost email, a previous version of a data file, or accidental user deletion of a critical file. As with their physical counterparts, doing a recovery here should be fast, simple, and available. These are the “fast” backups you use to get that piece of data that was lost.
System backups are used for complete restores. Your server catches fire, the hard drives fail, that kind of thing. They are pictures of the entire machine, and thus more complete yet more difficult to work with.
With virtualization, you actually can have both types at both the host and guest level. Many think that virtualization makes backups easier, but it actually doesn’t – it gives you more options, but doesn’t ease the complexity. In fact, the diversity of options may actually make backups harder. There is much more to manage, particularly if you want to take backups at all levels.
The key to success is a Service Level Agreement (SLA). The SLA defines exactly what the expectations are on all sides. The SLA is the contract glue that binds the technical capabilities to the customer needs. These SLA definitions become even more important in a virtualized world, as it’s easy to make assumptions about what is the current status. Defining your backup strategy as it relates to the virtualized environments will make the implementation a success.
What is truly dangerous is making assumptions about your restore capabilities. Sidekick users learned this the hard way when they were told their data was gone. As a customer’s trusted advisor, it’s our responsibility to make sure they (and we) aren’t making those dangerous assumptions. Just because it’s virtualized – or in the cloud – doesn’t make the SLA any less important. You know what they say about assumptions. It makes something unpleasant out of you and me.