Some notes from AWS REInvent
- Kill dev/test instances regularly (elastic instances)
- Re-evaluate instance type regularly
- Use reserved instances
- Spot instances + autoscaling
- Go serverless: Lambda + API gateway + DynamoDB
- Use elasticache
- S3 static web hosting
- Use an appropriate DB
- Don't use RDBMS as a blob store, use S3 instead
- Autoscaling optimizations: Use 6x3 over 12x2
- Manual shutdown
- Tag instances and have a lambda function to shut down regularly
- Datapipeline / cloudwatch