The AWS Developers Podcast Evolving Lambda: from ephemeral compute to durable execution
Mar 4, 2026
Michael Gasch, Product Manager for Lambda Durable Functions at AWS, walks through the evolution of serverless and the need for native orchestration. He discusses checkpoint-replay, wait patterns like callback and condition, LLM orchestration, ECS coordination, and when to pick Durable Functions versus other options. He also covers tooling wins, surprising customer feedback, and what’s coming next.
AI Snips
Chapters
Books
Transcript
Episode notes
Use Lambda Managed Instances For Predictable High Scale
- Use Lambda Managed Instances when you need predictable capacity and existing procurement benefits while keeping the Lambda programming model.
- Michael says this fits enterprises that want reserved-fleet economics but serverless operation.
Tenant Isolation For Service Providers
- Tenant isolation addresses multi-tenant providers' need to map logical tenants to dedicated sandboxes inside shared AWS accounts.
- Michael explains it reduces provider custom work for routing and per-tenant security inside a shared account.
Durable Functions Adds Checkpoint Replay To Lambda
- Durable Functions is Lambda plus a new SDK that provides checkpoint-replay semantics so long-running, multi-step business logic becomes resilient without external orchestrators.
- Michael describes steps (checkpoints) that persist progress so subsequent invocations replay and return cached step results.
