Skip to main content
Welcome. This lesson explains core resilience concepts in AWS and how to apply them to real-world systems such as e-commerce websites. The emphasis is on preserving the customer experience when components fail — not just preventing outages, but avoiding degraded user journeys that cost revenue and trust. Consider a busy sale day: the site can be “up” but if checkout fails or orders are lost, customers still experience a failure. A resilient architecture protects availability, performance, and data integrity across localized failures (instance, AZ) and large-scale events (region outages). If we’re operating globally and one data center fails, the website
A presentation slide titled "The Challenge" with three colored icons labeled "24/7 Availability," "Handling Traffic Spikes," and "Data Integrity." Each icon is a rounded square (blue, orange, green) with a simple white line illustration representing each challenge.
has to continue to function seamlessly. Key resilience goals for e-commerce
  • 24/7 availability so customers can complete purchases any time of day.
  • Handle traffic spikes (e.g., Black Friday) without service degradation.
  • Preserve data integrity for order history, billing, and customer data.
Core resilience concepts to design for in AWS
  1. High availability — keep services reachable even if whole servers or AZs fail.
  2. Fault tolerance — enable replacement components to take over quickly after a component failure.
  3. Scalability — elastically scale capacity up and down to match demand without unnecessary cost.
  4. Disaster recovery (DR) — recover from region-level failures and large-scale disasters.
High availability and fault tolerance
  • High availability distributes workloads across multiple Availability Zones (Multi-AZ) and uses services with built-in redundancy, for example:
    • Multi-AZ RDS
    • DynamoDB Global Tables
    • Elastic Load Balancers with cross-zone load balancing
  • Fault tolerance targets single-component failures (failed EBS volume, NIC, or EC2 instance). Typical patterns:
    • Auto Scaling groups with health checks to automatically replace unhealthy instances
    • Stateless application tiers that can be scaled horizontally
    • Stateful services with synchronous replication or quorum mechanisms
Scalability and cost considerations
  • Scale to meet peak traffic and scale back to control costs. Common patterns:
    • Auto Scaling (EC2 Auto Scaling Groups)
    • Serverless (AWS Lambda) for event-driven workloads
    • Storage tiering (S3 lifecycle policies, S3 Intelligent-Tiering)
  • Reduce cold-start impact for serverless functions by warming strategies, provisioned concurrency, or pre-provisioning compute.
Disaster recovery (global)
  • DR prepares for region-wide outages. Techniques include:
    • Route 53 DNS failover
    • Cross-region backups and replication (S3 Cross-Region Replication, RDS cross-region read replicas)
    • AWS Global Accelerator for resilient global traffic management
    • Multi-region deployments (pilot light, warm standby, or active-active)
Route 53 and multi-region routing
  • AWS recommends distributing workloads across Availability Zones, and when necessary, across multiple regions.
  • Route 53 routing policies let you control traffic direction: failover, weighted, latency-based, geolocation, and more.
  • Use multi-region strategies for global failover and low-latency access; use Multi-AZ for localized resiliency within a region.
  • Example strategy: run active-active with a small percentage of traffic in a secondary region (e.g., 95/5 or 80/20) and shift weight during failover.
A diagram titled "Resilience Techniques in AWS for an E‑Commerce Website" showing users routed through Amazon Route 53 to multiple regions. Each region contains load balancers, EC2 instances spread across two availability zones and Auto Scaling Groups for redundancy and failover.
Design patterns for regional failover
  • Active-active: both regions serve live traffic; reduces failover time and improves global latency. Requires data replication (DynamoDB Global Tables, bi-directional replication) or conflict resolution.
  • Active-passive (warm or cold standby): the passive region has pre-provisioned or minimal resources; failover is usually faster than cold starts but slower than active-active.
  • Pilot light: minimal resources kept warm in the DR region; scale up during recovery.
  • Use traffic shifting (weighted routing) for gradual cutovers and canary failovers during tests.
Best practices (practical recommendations) Implementation tips
  • Enable cross-zone load balancing on ALBs to spread traffic evenly across AZs.
  • Use read replicas in other regions for read-heavy workloads and as part of DR plans.
  • Test backup and restore procedures regularly (automated recovery drills).
  • Automate infrastructure as code (CloudFormation, Terraform) to reproduce environments quickly.
  • Monitor metrics and set SLOs/SLAs with alerting tied to runbooks for escalation.
A slide titled "AWS Resilience in E‑Commerce – Best Practices" listing recommendations like Multi‑AZ RDS/DynamoDB/ElastiCache, Auto Scaling Groups and ALBs, S3 versioning/AWS Backup/RDS read replicas, Route 53 failover/AWS Global Accelerator, and regular DR testing. The slide includes an AWS logo on the left and a shopping/checkout icon on the right.
Plan and test your recovery strategies proactively. Regular DR drills (pilot light, warm standby, active-active) reveal gaps in automation, runbooks, and replication before a real incident occurs. Combine automated tests with manual runbooks to validate both automatic and human-driven recovery steps.
Testing and validation
  • Schedule regular DR exercises and validate RTO/RPO against business requirements.
  • Use traffic shifting (Route 53 weighted routing) to perform canary failovers.
  • Verify data consistency after failover (check replication lag, reconciliation steps).
  • Maintain runbooks and automate as much of the failover process as possible.
Summary This guide outlined primary AWS resilience patterns to protect customer experience, control costs, and meet SLAs. The recommended approach combines Multi-AZ and multi-region deployment patterns, automated scaling and recovery, and a disciplined testing program to validate assumptions before an outage occurs. Links and references

Watch Video