AWS re:Invent 2025 - Designing resilient Serverless Applications (API311)

Designing Resilient Serverless Applications

Importance of Resiliency

• Resiliency is crucial for serverless applications to avoid costly message loss and downtime
• Financial and brand impact of unreliable systems highlighted by Gartner research
• Shared responsibility model for resiliency - AWS manages infrastructure, developers responsible for application-level resiliency

Serverless Resiliency Challenges

• Potential issues include service limits, misconfigurations, resource availability, and state management
• Avoiding anti-patterns like excessive function chaining is key

AWS Well-Architected Framework and Serverless Lens

• Well-Architected Framework's reliability pillar is the focus for resiliency
• Serverless Lens provides deeper guidance on application-level resiliency best practices
• Power Tools for AWS codifies these best practices across multiple languages

Serverless Resiliency Features

API Gateway

• Throttling at account, stage, and method levels to regulate inbound traffic
• Method-level caching and CloudFront integration for reducing backend load

Lambda

• Synchronous vs asynchronous invocation models have different resiliency characteristics
• Asynchronous invocations leverage internal queuing and retries for improved reliability
• Event source mappings (e.g. SQS, DynamoDB Streams) provide additional resiliency options

EventBridge

• Automatic retries with exponential backoff for event delivery
• Ability to configure dead-letter queues and delivery windows

Power Tools for AWS

• Provides utilities for batch processing, item deduplication, and other resiliency patterns
• Simplifies implementation of best practices across languages

Resiliency Patterns

API Gateway to SQS/Step Functions

• Bypassing Lambda to directly integrate API Gateway with queues or workflows
• Decouples request handling from processing logic

Saga Pattern for Distributed Transactions

• Leverages compensating actions to maintain consistency across multiple steps
• Crucial for long-running, multi-service business processes

Validating Resiliency

Chaos Engineering with Fault Injection Service

• Ability to inject failures, latency, and other disruptions to test system behavior
• Helps identify weaknesses and validate recovery mechanisms

Key Takeaways

• Every message and request is important - can't afford to lose the "million-dollar widget"
• Carefully consider service interactions and assumptions when designing architectures
• Leverage serverless resiliency features and patterns to build more reliable applications
• Validate resiliency through chaos engineering and fault injection testing