With the multiple availability zone concept, end users can deploy their applications in a distributed model, so when one availability zone fails, the applications continue running. The end user is responsible for maintaining the application state between zones. The provider must have two sites (data centers) distant enough to decorrelate their probability of failure, but close enough to have them under the same cloud management. The provider must also secure disaster recovery of the cloud management components.

There are two options to design multiple availability zones, and they differ as to how resource group management components are deployed – in distributed or stretched configurations.