By Stacy Griggs

Last week RackSpace had an outage that impacted their cloud, Amazon (EC2) had a similar outage at their Virginia datacenter earlier this month. Twitter, Facebook and the blogosphere are filled with predictions that these events are a black eye for cloud computing because sites like Brizzley and TechCrunch were down. The real reason those sites were offline has little to do with cloud computing. These sites were down because their cloud platform was located in single datacenter (Note – it is not uncommon for providers to host all of a clients cloud instances in single datacenter). When cloud providers offer support for multiple datacenters, clients have to expressly elect this option and pay extra for it. For example in the Amazon Cloud you can purchase additional instances in different availability zones. You are charged for the data transfer between instances plus the cost of the redundant instances, which can more than double the cost of your solution.

The reason many public clouds are hosted in a single facility is simple, there are significant cost involved with implementing global server load balancing and maintaining a hot secondary site. Early adaptors of the cloud (eg not enterprise customers) were often not willing to pay for this level of redundancy. While most cloud environments are highly available because of their lack of dependence on single hardware nodes they continue to be vulnerable to a failure of power or loss of connectivity at the facility in which they are housed. As enterprise customers have migrated to cloud computing environments, this has been part of the cloud that they understand least.