When even higher availability is required to prevent an outage due to an AWS region failure, then we have to choose multiple regions to host our application. We can choose from a number of regions, and these have a geographical distribution that aligns well with the developed world's population density. The following map outlines the AWS regions that are currently in operation or under construction and will be online in the near future:

If we want to deploy our application across multiple regions, we need to consider the impact of such a decision. Within a region, the connections between datacenters and AZs are low latency, and are usually layer 2 links. Across regions, traffic is carried across a long-distance public Internet infrastructure. The following considerations need to be accounted for:

• Latency: We won't be able to maintain synchronous replication across multiple regions, since the latency between regions will be counted in tens to hundreds of milliseconds
• Cost: Since the traffic being sent out of a region traverses the internet, it will be counted as standard outgoing traffic and billed accordingly
• Lack of built-in redundancy: Only a selection of services are designed to provide built-in replication of data to multiple regions
• Recovery from region failure: You will need to design recovery and re-synchronization procedures for services that are deployed across multiple regions
• Latency to your clients: Choosing a region far from your clients or failing over to such a region will have an impact on the latency that's experienced by your end users