On October 20th, a major AWS outage in the us-east-1 region severely impacted services, including DynamoDB, which faced increased error rates. This incident affected high-profile companies like Disney+, Lyft, and Reddit. The author emphasizes that while their infrastructure relies on AWS, they must support customers who choose to operate in us-east-1, despite its vulnerabilities. Acknowledging the critical nature of their service, which provides authentication and access control, the author details the importance of maintaining high reliability, aiming for a five nines SLA, equating to just over five minutes of downtime per year. The article delves into the challenges of achieving such reliability within a cloud infrastructure that experiences frequent outages. Historical incidents, including hardware failures, human errors, and natural disasters, highlight the inherent risks of relying solely on AWS services. The author argues against depending on AWS’s own SLAs, which fall short of five nines in several key services like Lambda and API Gateway. The core message is clear: to meet strict reliability commitments, they must build resilience into their architecture, rather than relying on the cloud provider's assurances.