This paper describes how Amazon Web Services (AWS) Lambda addresses the challenge of loading large container images quickly to support serverless applications.

AWS Lambda uses a system that combines several techniques, including caching, deduplication, convergent encryption, erasure coding, and block-level demand loading. The paper describes how these techniques work together to achieve high scalability and low cold-start times. It also covers some challenges faced in implementing the system, such as ensuring data security and handling metadata.

Finally, it presents the system’s performance characteristics and the authors' experiences with deploying it in production.

https://drive.google.com/file/d/1MxqRzErd3Vpti_mHkINpVzTrX-bzBlZl/view