Who backs up the cloud when the cloud goes down?

One of Amazon’s Sydney Data Centres went down over the weekend thanks to a severe storm.  It shut down websites for some of Australia’s biggest web businesses.  Domain, The Iconic and Domino’s were severely affected and Carsales plus REA Group were affected to a lesser extent.

One of the biggest cloud companies, AWS is built around the concept of reliability and availability, but even it couldn’t manage to keep services uninterrupted when a chained series of events prevented failover strategies from effective execution.

It makes the case for owning the removal of geographic risk.  If your company relies on access to data, either internally or externally, then you need to have a second site for your data, running in parallel or available in a matter of minutes should you first location go down.  But the network components that will manage a failover shouldn’t be located in the same primary data centre or the second site could be rendered ineffective, such as what happened in this instance.

REA Group delivered service uninterrupted because it deploys to two of Amazon’s Availability Zones in two different regions simultaneously, so wasn’t impacted by the API difficulties that caught the other brands out.

For smaller businesses with a reliance on data availability, the implications are clear.  The cloud is not a panacea.  You cannot offload your responsibilities just by signing up.  Smart architecture is still required, geographic risk needs to be mitigated as much as any other kind, and regular disaster recovery testing is the only way to prepare for the unexpected.

James Walker