An advantage of using Amazon Elastic Compute Cloud (EC2) is the ability to start, stop, create, and terminate instances at any time. However, this flexibility creates a potential challenge with IP addresses. Restarting a stopped instance (or re-creating an instance after another instance is terminated) results in a new IP address. How do you successfully reference a machine when the IP address is constantly changing?
In response to this problem, Amazon offers the ability to allocate an Elastic IP address. An Elastic IP provides you a single IP address that you can associate with different EC2 instances over time. If your Amazon Elastic Compute Cloud-Virtual Private Cloud (VPC) or Amazon Elastic Compute Cloud-Classic instance has an Elastic IP and that instance is ever stopped or terminated, you can immediately associate a new EC2 instance with the Elastic IP. Your existing applications will not break because the applications see the IP address they were expecting, even though the back-end EC2 instance has changed.
If you build your site using the Amazon Web Services (AWS) Management Console, you can associate an Elastic IP with your stand-alone ArcGIS Server site or single-machine ArcGIS Enterprise deployment if you are not already using an Amazon Load Balancer.
If you ever need to stop an instance, you should reassociate it with its Elastic IP after you start the instance again. You can even associate the Elastic IP with a backup instance while the other instance is down. If you don't have an Elastic IP, users' connections to your machines will permanently break if you ever have to stop the instance.
An Elastic IP is not the same as an Amazon Load Balancer. An Amazon Load Balancer helps you scale out your site by associating many EC2 instances at the same time under one web address. An Elastic IP, on the other hand, can only be associated with one EC2 instance at a time.
Follow the instructions in AWS documentation to allocate an Elastic IP address and associate it with an AWS instance.