1.1 Why Deployment Strategies Matter
The way you deploy new application versions directly affects availability, risk, rollback speed, and cost. AWS provides multiple deployment strategies across different services. Choosing the right strategy depends on your tolerance for downtime, risk appetite, and infrastructure type.
1.2 Complete Deployment Strategies
All-at-Once (Big Bang)
- Fastest deployment. Simplest to implement.
- Downtime during update. Not suitable for production.
- Use for: dev/test environments, non-critical workloads
Rolling
- No downtime but reduced capacity during rollout
- Two versions run simultaneously (must handle compatibility)
- Rollback is slow (must roll forward or re-deploy previous version)
- Use for: when you can tolerate reduced capacity and mixed versions
Rolling with Additional Batch
- Maintains full capacity throughout deployment
- Extra instances launched temporarily (slightly higher cost during deploy)
- Use for: production workloads that cannot lose capacity
Immutable
- Creates brand new instances (never modifies existing)
- Full capacity throughout. Quick rollback (terminate new instances).
- Higher cost during deployment (double infrastructure temporarily)
- Use for: when you want guaranteed clean deployments, quick rollback
Blue/Green
- Two identical environments. Swap traffic via DNS (Route 53) or ALB
- Zero downtime. Instant rollback (redirect traffic back to Blue)
- Most expensive (double infrastructure during deploy, possibly longer)
- Use for: critical production systems requiring zero downtime + instant rollback
Canary
- Deploy to a small subset first, validate with real traffic
- Gradually shift traffic if healthy. Instant rollback if not.
- Requires good monitoring and automated rollback triggers
- Use for: high-risk changes, production safety, regulated environments
A/B Testing (Traffic Splitting)
- Similar to Canary but driven by business metrics, not just health
- Route specific user segments to different versions (by header, cookie, geography)
- Measure business outcomes: conversion rate, revenue, engagement
- Use Route 53 Weighted or ALB Weighted Target Groups
- Use for: feature experiments, UX testing, business-driven decisions