5 AWS mistakes you should avoid

The post says typical AWS web applications should use a load balancer, a scalable web backend, and a database. It then lists five recurring mistakes the author sees in small and medium AWS deployments and says they should be avoided.

Reading notes

#

• Infrastructure should not be managed manually through the console; CloudFormation is presented as the way to describe resources in a template and create or update a stack reproducibly.
• Every EC2 instance should run inside an Auto Scaling Group, even a single instance, because the group monitors the instance and can later support scaling based on alarms.
• CloudWatch metrics should be analyzed, not just collected; the example shows a daily CPU spike caused by a cron job, and alarms should be defined after the metrics are understood.
• Trusted Advisor checks an AWS account against AWS best practices in cost optimization, performance, security, and fault tolerance, and the post recommends paying attention to security first.
• Virtual machines should be downsized when underutilized, using CloudWatch metrics to decide whether instance size or fleet size can be reduced.