Cold Start Optimization Techniques for Function‐as‐a‐Service Using Lambda@Edge and Container Reuse

This chapter examines new optimization policies to reduce cold start penalties by implementing Lambda@Edge features and container reuse policies, over an 18-month time span in various regions in the AWS, where more than 15 million monthly requests are serviced. The recent developments in the area of serverless computing emphasize its increasing value in terms of streamlining distributed, scalable and intelligent cloud-based systems. Container reuse policies were also implemented via globalization of variable initialization patterns and common usage of libraries, which allowed pooling connections, caching and maintenance of state across each invocation in warm containers. Container reuse strategies demonstrated substantial improvements in warm invocation performance through connection pooling and in-memory caching mechanisms. Reuse of connection pools alone decreased the mean latency by 37.2%, mostly due to the removal of connection establishment overheads per invocation to the database.