Opinion

posted on 29 August 2008 11:11

Enterprises have known this inconvenient truth for years and years: if an IT resource is critical then it needs an offsite disaster recovery facility. The would-be cloud storage behemoths thought they didn't need offsite DR. We've got news for them - they do.

Amazon's S3 service went down. Apple's MobileMe service went down. Google's cloud service went down. Now XCalibre's FlexiScale service has gone down - because an engineer adding capacity to the system deleted a storage volume. A classic unanticipated human error.

Are the people running these cloud service operations recruited from the pool of failed UK government IT directors? (An expanding pool of proven idiots adept at losing vast amounts of personal data and interrupting service delivery.)

Each of these services think that because they have distributed data centres they have a DR capability. They don't. What they have is a single cloud service capability in which catastrophic errors cause the shutting down of the whole system. This is not disaster recovery - it's disaster continuance.

What they need is a completely separate cloud with a heartbeat facility and failover, a global cloud-scale version of DoubleTake or NeverFail server replication - cloud replication. That would cost big, big bucks but it would ensure that if cloud one crashed, cloud two could keep raining service onto users.

It's insurance, just like big insurance companies who insure their own risks with separate re-insurance companies.

So long as cloud service providers have no separate backup cloud systems then the potential for cloud crashes remains and at some time their clouds will crash.

^{[Chris Mellor.]}