Скачать или смотреть Auto Scaling AWS EC2 Instances Made Easy | Autoscaling Groups & Launch Templates

AWS EC2 Auto Scaling is a service that automatically adjusts the number of Amazon EC2 instances in your application environment to maintain performance and reduce costs. It helps ensure you have the right number of instances running at the right time—scaling out (adding instances) when demand increases and in (removing instances) when demand drops.

Why Auto Scaling Is Useful
Resilience: It maintains application availability during sudden spikes in traffic.

Cost Efficiency: It scales in during low demand, so you're not paying for unused capacity.

Automation: It reduces manual intervention by responding automatically to changes in demand.

High Availability: It can detect unhealthy instances and replace them automatically.

How Launch Templates Work
A launch template defines the configuration used when EC2 instances are launched. It includes settings such as:

AMI ID (Amazon Machine Image)

Instance type (e.g., t3.micro, m5.large)

Key pair

Security groups

User data (e.g., bootstrapping scripts)

Storage volumes

IAM role

Launch templates help standardize instance creation and are a prerequisite for EC2 Auto Scaling groups. You can version them to manage updates to your application easily.

How Auto Scaling Groups Work
An Auto Scaling Group (ASG) is a collection of EC2 instances managed collectively based on specified scaling policies. It includes:

Launch template: Defines what kind of instance to start.

Min, Max, Desired capacity: Defines the lower and upper bounds and the target number of instances.

Scaling policies: Define how and when to scale in or out.

Health checks: Replace failed instances automatically.

Availability Zones: Spread instances across multiple AZs for resilience.

Provisioning Strategy
When provisioning, the ASG chooses which Availability Zones to place instances in. It follows your scaling policies and health check configurations. If capacity is unavailable in one zone, it attempts another.

You can also configure "launch then terminate" behavior when changing configurations. This tells the ASG to first launch a new instance (e.g., from a new launch template version), then gracefully terminate the old one. This avoids temporary under-capacity during updates.

Scaling Metrics and Policies
Auto Scaling supports dynamic and scheduled scaling. The most common metrics include:

CPU Utilization (e.g., scale out if over 70 percent for 5 minutes)

Network traffic (e.g., bytes in/out)

Custom CloudWatch metrics (e.g., queue length from SQS, request count)

Application Load Balancer request count per target

You can use:

Target tracking: Automatically maintain a target metric (e.g., 50 percent average CPU).

Step scaling: Add or remove instances in proportion to how far the metric deviates.

Scheduled scaling: Define exact times (e.g., scale up weekdays at 8 AM).

How Scaling Down Works
Scaling in works by:

Identifying underutilized instances (based on metrics).

Terminating instances with the least impact, often the oldest or least recently used.

Respecting cooldown periods to avoid too frequent changes.

Gracefully draining connections if tied to load balancers.

Auto Scaling tries to maintain balance across Availability Zones and avoids removing instances that are actively handling requests unless necessary.