AWS EC2: Instances, AMIs, Security Groups, and Pricing

# Core concepts:
# AMI (Amazon Machine Image)  — OS + pre-installed software template
# Instance type               — CPU, memory, storage, networking specs
# Key pair                    — SSH key for secure access
# Security group              — virtual firewall controlling traffic
# Elastic IP                  — static public IP address

# Instance lifecycle:
# Pending → Running → Stopping → Stopped → Terminated
#
# Stopped vs Terminated:
# Stopped: instance is off, EBS data persists, still billed for storage
# Terminated: instance deleted, EBS deleted (unless configured otherwise)

# General Purpose    — balanced compute/memory/network
# t3, t4g             — burstable (good for dev/test, variable load)
# m5, m6i             — steady workloads

# Compute Optimised   — high CPU, lower memory
# c5, c6i             — batch processing, HPC, gaming servers

# Memory Optimised    — high memory/CPU ratio
# r5, r6i, x1         — in-memory databases, real-time processing

# Storage Optimised   — high sequential I/O
# i3, d2              — data warehousing, distributed file systems

# Accelerated Computing
# p3, p4, g4dn        — GPU instances: ML training, video rendering

# Naming convention: m5.xlarge
# m = family, 5 = generation, xlarge = size

# Security group rules:
# Inbound rules: traffic ALLOWED INTO the instance
# Outbound rules: traffic ALLOWED OUT of the instance (default: all allowed)

# Rules specify:
# - Type (SSH, HTTP, HTTPS, custom)
# - Protocol (TCP, UDP, ICMP)
# - Port range (22, 80, 443, or range like 8000-9000)
# - Source (IP CIDR, security group ID)

# Example: web server security group
# Inbound: HTTP (80) from 0.0.0.0/0 (anywhere)
# Inbound: HTTPS (443) from 0.0.0.0/0
# Inbound: SSH (22) from your-office-IP/32 only
# Outbound: All traffic (default)

# Key properties:
# - Stateful: if inbound is allowed, response is automatically allowed
# - Default: deny all inbound, allow all outbound

# On-Demand
# - Pay by the hour/second, no commitment
# - Most expensive, maximum flexibility
# - Use: unpredictable workloads, dev/test

# Reserved Instances (RI)
# - 1 or 3 year commitment
# - Up to 72% cheaper than On-Demand
# - Use: steady-state, predictable workloads

# Spot Instances
# - Use spare EC2 capacity
# - Up to 90% cheaper than On-Demand
# - Can be interrupted with 2-minute notice
# - Use: batch jobs, data processing, fault-tolerant workloads

# Savings Plans
# - Flexible alternative to RIs
# - Commit to $/hour spend, not specific instance type
# - Applies to EC2, Lambda, and Fargate

# Dedicated Host
# - Physical server dedicated to you
# - Use: compliance requirements, bring-your-own-license

Exam tipThe most common EC2 exam question: "Which pricing model for a steady-state, 24/7 web server?" — Reserved Instances (1 or 3 year). For unpredictable/batch: On-Demand or Spot. For compliance: Dedicated Hosts.