Redundancy provides multiple interchangeable components or paths so one failure does not interrupt service. In a web application, running several EC2 instances instead of one removes a single point of failure, especially when the components are decoupled. section 16.0
A redundant AWS design distributes both workloads and their supporting infrastructure:
- Run application instances across multiple Availability Zones, with a subnet in each zone, so an outage in one zone leaves capacity in another. section 16.1.2
- Place an Elastic Load Balancer between clients and instances so requests can reach surviving instances and avoid failed ones. section 16.1.1 section 16.1.2
- Use an Auto Scaling group to replace a crashed EC2 instance, or use SQS to decouple message producers from redundant consumers. section 16.1.2
Three production lines
A single pie-production line stops completely when one step fails. Three independent lines can continue producing when one line fails, although they require more machines. The same pattern applies to EC2: three smaller instances can provide redundancy instead of one large instance, with a load balancer directing incoming requests. section 16.1.1
Shows how an Auto Scaling group and Elastic Load Balancer combine with redundant EC2 instances to preserve service after an instance failure. section 16.1.2
