By default, all Amazon EC2 instances are assigned two IP addresses at launch: a private (RFC 1918) address and a public address that is mapped to the private IP address through Network Address Translation (NAT).

If you use dynamic DNS to map an existing DNS name to a new instance's public IP address, it might take up to 24 hours for the IP address to propagate through the Internet. As a result, new instances might not receive traffic while terminated instances continue to receive requests.

To solve this problem, Amazon EC2 provides elastic IP addresses. Elastic IP addresses are static IP addresses designed for dynamic cloud computing. Elastic IP addresses are associated with your account, not specific instances. Any elastic IP addresses that you associate with your account remain associated with your account until you explicitly release them. Unlike traditional static IP addresses, however, elastic IP addresses allow you to mask instance or Availability Zone failures by rapidly remapping your public IP addresses to any instance in your account.

In the following image, web servers are connected to the Internet through elastic IP addresses and to database servers through their private IP addresses.

The administrator decides to replace a web server with a larger instance type. To do this, the administrator starts a new instance using a larger instance type (1), disassociates an elastic IP address from a running instance (2), associates the elastic IP address with the new instance (3), and terminates the old instance (4).