A little news on backups and why they matter

One of our sysadmins has been cleaning up and auditing our backups recently, and we thought we’d look at some of the numbers involved and how they’ve changed over time.

Several years ago we used tapes exclusively, and they were always taken offsite. We don’t have the numbers from that time, but every fully-managed server came with backups included. AMANDA is still our backup software of choice and it’s been a reliable system for many years.

That said, tapes aren’t really a good match for many of our customers’ needs, and the costs can be prohibitive. That’s why we developed our onsite disk-based backup solution. As well as being able to offer large amounts of storage at a much lower cost to the customer, the experience is actually better because you don’t have to wait for backup tapes to be shipped to the datacentre in the event of a recovery.

Onsite backups saw a great uptake and they’ve been sold with many new servers since then. Of course with such a difference in the backup offerings, there has to be a tradeoff. Namely, that your backups are no longer offsite.

To be frank, this is absolutely fine for a vast majority of customers. Backups are an overhead cost no matter how you look at them, so it’s important to ask yourself just what you need to protect against when choosing backups.

  • Offsite backups: frequently performed with magnetic tape, but also possible with online storage, offsite backups tend to be an enterprise-level requirement.
    As well as protecting from accidental deletion or corruption, offsite backups guard against destruction of the facility (eg. fire, earthquake) and are considered mandatory for critical data that can’t afford to be lost.
    Depending on your industry, offsite/offline backups may be a legal requirement.
  • Onsite backups: onsite (and implicitly online) backups allow for data recovery in the face of deletion and corruption. This represents the most common threat by far, and is the suitable choice in most cases.
    Unless the primary server is completely wiped out, recovery is typically fast and easy. If worst comes to worst, provisioning may take a long time and keep you offline, but your data is safe.
  • No backups: almost universally a bad idea unless you really know that there’s no important data on your server, and there never will be.
    Large deployments with many servers (10-20) tend do this, but only because there’s a clear separation of roles (eg. the webservers don’t need backups because all the data lives on the databases and fileservers, which are backed up).

It’s this last category that worries us the most. Most of us have bigger fish to fry if a meteorite hits the datacentre and and makes a new crater-shaped home for our server, but the threat of losing all of your data due to hardware failure or human error with no prospect of recovery is very real.

Emails, customer data, sales records, financial details – sudden loss of even one of these would badly cripple most businesses, and we know that in many cases they’re centralised onto one server. It’s hard to overstate: Having backups is really important!

With this in mind, we decided to do a quick headcount while auditing our backups. Some of Anchor’s many servers are part of the “large deployments” that we mentioned, but with the numbers we’re looking at they shouldn’t skew things too badly.

We found that 72% of our deployed servers have backups, and about 6.5% of those are offsite backups. Those are pretty good numbers, as we know that a lot of our own infrastructure doesn’t need backups. For the rest of them, we’re planning on working to improve the coverage.

If you’re not absolutely certain whether your own server has backups, find out! The worst time to find out is after you’ve lost data. Backups are cheap to have and easy to get setup, so there’s almost no excuse to not have them.