Elastic Load Balancing and EC2 Auto Scaling in twenty-nine slides: a load-balanced web tier and the load balancer's nodes; the Application, Network and Gateway Load Balancers, and what an ALB and an NLB each see; listeners and rules, target groups, deregistration and health checks; sticky sessions, cross-zone load balancing and TLS termination; launch templates, Auto Scaling groups, replacing an unhealthy instance and the termination policy; target tracking, step, simple, scheduled and predictive scaling, warm-up, cooldown, lifecycle hooks and which policy fits which traffic; stateless servers and where their state goes. Every section ends with five exam questions.
The slides
The tool is twenty-nine slides: an introduction, then six sections, listed on its left. Every section ends with five
exam questions, each with its answer and the tempting wrong answer.
- 0. Introduction
- A load-balanced web tier: the ALB, the Auto Scaling group, the launch template and the alarm.
- Elastic Load Balancing: one DNS name, a node in each AZ.
- 1. Load balancers
- The three load balancers: Application, Network and Gateway, side by side.
- ALB and NLB: what each one sees: two connections or one,
X-Forwarded-For. - The Gateway Load Balancer: traffic sent through firewall appliances.
- 2. Listeners, targets and health checks
- Listeners and rules: conditions, actions and priorities.
- Target groups: the target types, and deregistration delay.
- Health checks: the defaults, and a target's states.
- 3. Sessions, zones and TLS
- Sticky sessions: the
AWSALB cookie. - Cross-zone load balancing: two targets in one AZ, eight in the other.
- TLS termination: at the load balancer, encrypted again, or passed through.
- 4. Auto Scaling groups
- A launch template: versions, and the project's template.
- An Auto Scaling group: minimum, desired and maximum, subnets, target group, health checks.
- Replacing an unhealthy instance: step by step.
- Which instance a scale-in removes: the default termination policy, and scale-in protection.
- 5. Scaling policies
- Target tracking scaling: a day of load, the group following it.
- Step and simple scaling: steps by how far the alarm is breached.
- Scheduled and predictive scaling: a calendar, and a forecast.
- Warm-up, cooldown and the grace period.
- Lifecycle hooks: the two wait states.
- Matching a scaling policy to the traffic: the exam's table.
- 6. Stateless design
- Stateless servers: state on the server, and state in shared services.
- Where each kind of state goes: sessions, files, data, configuration, logs.
The buttons on the left and right edges move between slides. So do the arrow keys and a swipe.
The lab
Load Balancing Lab: the ALB and the Auto Scaling Group practices it.
The older
ALB Walkthrough creates an ALB step by step in the console.
The load balancers' protocols and target types, the health check defaults, the deregistration delay, the cross-zone
defaults, the stickiness limits, the grace period, the cooldown, predictive scaling's history and forecast, the lifecycle
hooks' timeout and the default termination policy were read from AWS's documentation on 2026-11-16. The charts are drawn
from made-up loads. The IDs are placeholders in AWS's form, and the addresses come from the documentation ranges.
Sources