On March 24, online encyclopedia giant Wikipedia went offline for more than 2 hours because of an overheating problem in one of their data centers. Even though they had had a DNS failover procedure, it was broken. The result: millions of users could not access Wikipedia for hours.

This situation is unfortunate for Wikipedia, but it would be even more unfortunate if you were an online business that lost 2 hours of critical revenue.

It is not enough to simply have a single back-up IP address and a manual switch to flip as a failover process.

Wikipedia’s outage serves to illustrate the need for not just a failover process, but an integrated approach to monitoring and IP failover services that ensures your Web sites will always be up and running as expected.

Our tips for best practices are:

• Have continuous monitoring in place for your DNS and also your Web servers (http and https) to make sure that they are always responding and available.

• Use a service that can automate the monitoring of specific content or images that should be served on these sites so you know they are not just up, but also working properly.

• Have more than one IP address, ideally located in different data centers that are geographically dispersed, where back-up copies of your Web site reside.

• Ensure that you use a failover service that combines the monitoring process with automatically failing over to a list of your back-up IPs to create a seamless failover process that requires no manual intervention.

Of course we also recommend that you ensure that traffic to your Web site is being managed on a reliable DNS network that provides similar monitoring and failover services. With a professional managed DNS provider, generally you will get the benefits of a globally distributed Anycast network that seamlessly handles load dispersion and failover, with higher capacity than you could reasonably build in house.

Written by John L. Kane, Vice President, Corporate Services at Afilias.