f you are tuned into web hosting news, then you know that ThePlanet had a fire in their data center earlier this month. We covered the outage in our blog posts as we helped clients with recovery efforts. ThePlanet’s service outage underscores the need for disaster recovery planning. A key decision in any disaster recovery plan is where to locate your backups.
Backup Location
At the height of the outage, we had some 50+ servers impacted. Fortunately, most were on the second floor or only impacted by DNS. Others, however, were on the first floor. Some clients did have backups available but these backups were also located in H1. I never really placed a huge importance on backup location selection for non-critical operations until the meltdown at ThePlanet. After that incident, backup location became something I began to reconsider.
On-site vs. Off-site
Essentially, the choice for backups is between an on-site or same facility location and an off-site location. Each location has benefits and drawbacks. Of course, keeping both on and off-site backups is ideal, but IT budgets may not permit such a robust solution. If you can only have one backup location, which should it be?
Here are two key decision points. I am sure there are others but these are two that likely dictate what location you select:
Critical Data
Critical data should always be backed up to a location that is not likely to be impacted by events at the primary location. I also suggest you assure that you have two independent network routes to the data. This way an event in one location has a slim chance of impacting data and a secondary site. We have some clients that backup data to our systems in Dallas and in NYC. These are geographically isolated locations on independent networks.
Recovery Time
The problem with remote, off-site backups is that for large datasets it can take significant time to move the data back to your primary location. Unless you are using Global Server Load Balancing or plan to switch DNS, having large sized backups at a remote location could mean hours of downtime.
A 10 megabit link will only transfer around 4 GB per hour. You may see higher throughput on some links but rarely will you see more than 20-30Mbit. If you have 10’s or 100’s of GB of data to move, the time to recovery could be days not hours.
If you need rapid recovery time, you may have to consider having both on-site and off-site backups. The off-site backups serve are there for major events while the on-site backups are there for minor events.
Data Responsibility
Disasters do happen. Apparently, some servers did have critical data loss as a result of the issues at ThePlanet. They offered to pay for data recovery services, though they had no obligation to do so. What if your data could not be recovered? How would that impact your business? Do your clients expect you to have backups? The impact of data loss to your business may be the key factor in selecting a backup location and solution.
Types of Failures
Failures can come from many angles. Human error, hardware failure, and natural disasters. What if a flood, hurricane or tornado rendered your server useless? What if fire destroyed the facility? Though unlikely, these failures must be considered when looking at overall risk. I suspect many readers of this blog may only have one copy of their data. What if that copy was lost? How much time, effort and money would it take to recover?
If I had to select just one….
If my IT budget only permitted one set of backups, I would select an off-site location. While it may hamper recovery time, an off-site location will protect against the widest array of issues. In a couple of weeks, we will be introducing some new backup services. The primary goal is to get our clients (and hopefully new ones) thinking about disaster recovery. One of the first steps is putting in place reliable backups at a strategically chosen location.