Return to Well-Architected Framework Guide

Appendix: Reliability

How do you design your workload service architecture?

• Choose how to segment your workload
• Build services focused on specific business domains and functionality
• Provide service contracts per API

How do you design interactions in a distributed system to prevent failures?

• Identify which kind of distributed system is required
• Implement loosely coupled dependencies
• Make all responses idempotent
• Do constant work

How do you design interactions in a distributed system to mitigate or withstand failures?

• Implement graceful degradation to transform applicable hard dependencies into soft dependencies
• Throttle requests
• Control and limit retry calls
• Fail fast and limit queues
• Set client timeouts
• Make services stateless where possible
• Implement emergency levers

Return to Well-Architected Framework Guide