In this subchapter, we look at two EC2 features that solve specific problems: Elastic IPs (to have a fixed IP address) and Placement Groups (to control where your instances are physically placed). You’ll use the first one very soon; the second is for more specialized cases, but it’s good to know about it.

Each EC2 instance has IP addresses to communicate over the network:

• Private IP: to talk to other resources within your network (VPC). It remains as long as the instance exists.
• Public IP: so the internet can access it. Here’s the problem: by default, this public IP changes every time you stop and start the instance.

Why this is a problem: Imagine you point your website’s name (myweb.com) to the public IP 54.12.34.56. You stop the instance at night to save money (subchapter 4.3) and, when you start it in the morning, AWS gives it another IP, for example 54.99.88.77. Now your website “doesn’t work” because the name still points to the old IP.

An Elastic IP (EIP) is a fixed public IP address that you reserve and can associate with an instance. Unlike a normal public IP, it doesn’t change, even if you stop and start the instance.

Analogy: A normal public IP is like a phone number that gets reassigned to you each time. An Elastic IP is like buying your own fixed number: it’s yours and you’re always reachable at it.

Advantages of an Elastic IP:

• It’s constant: ideal for pointing domain names.
• You can move it from one instance to another. If an instance fails, you associate the EIP to a backup and traffic keeps arriving at the same place. This is useful for failover.

⚠️ Cost detail (important): AWS charges you for Elastic IPs that you reserve but DO NOT use (those not associated with a running instance). The logic: IPv4 addresses are a scarce resource, so AWS penalizes hoarding them without use. Release Elastic IPs you don’t need to avoid unnecessary charges. Since 2024, AWS also charges for public IPv4s in general, so use them wisely.

Modern note: In current architectures, Elastic IPs are often not used directly. Instead, a load balancer (Chapter 13) or CloudFront (Chapter 16) is placed in front, providing a stable entry point without tying an IP to a specific instance. Still, you should know about EIPs because they’re useful in many cases.

By default, AWS decides where to physically place your instances within a datacenter. Normally that’s fine. But sometimes you want to influence that placement for performance or resilience reasons. That’s what Placement Groups are for.

There are three strategies, each solving a different need:

Places the instances very close together, in the same physical location, so they communicate with lowest latency and highest bandwidth possible.

• Use case: applications requiring ultra-fast communication between instances (high-performance computing, big data, scientific calculations).
• Risk: since they’re all together, a physical failure could affect them all at once. Less resilience in exchange for more speed.

Places each instance on different physical hardware, as far apart as possible.

• Use case: critical applications where you don’t want two instances sharing the same point of failure. If a physical server dies, only one instance goes down.
• Example: the few servers supporting a critical system, where losing two at once would be unacceptable.

Divides instances into groups (partitions), each on separate hardware. Instances within a partition share infrastructure, but partitions themselves are isolated from each other.

• Use case: large distributed systems (like Hadoop, Cassandra, Kafka databases) that already manage replicas and want to control which groups can fail together.

Don’t worry if Placement Groups seem advanced: they are. Most projects don’t need them at first. What’s important now is to know that they exist and what each strategy is for.

• The normal public IP of an instance changes when stopping/starting; this breaks domains pointing to it.
• An Elastic IP is a fixed public IP that you reserve and can move between instances (useful for failover).
• Watch out for costs: AWS charges for reserved and unused Elastic IPs. Release those you don’t need.
• In modern architectures, a load balancer or CloudFront usually replaces the Elastic IP as a stable entry point.
• Placement Groups control physical placement: Cluster (performance), Spread (resilience), and Partition (large-scale balance). They’re for advanced cases.

In the last EC2 subchapter, we’ll look at something that directly affects your bill: purchase models (On-Demand, Reserved, Savings Plans, and Spot) and when to use each one.