PARSEC: an adaptive and efficient platform for reducing cold start in serverless computing

Abstract

Serverless computing has gained significant attention due to its cost-effectiveness and scalability. Despite its advantages, serverless computing faces the critical issue of cold starts – delays encountered during the initial invocation of idle functions. This research investigates various strategies to reduce these delays, focusing on improved sandboxing, efficient package management, and optimized function startup. A key part of the study is understanding the technological aspects that contribute to cold starts, including optimizing the virtualization layer and effective container provisioning. Using the OpenLambda framework, the research delves into the complexities of serverless architectures, seeking innovative solutions to enhance performance. The findings of this study, encapsulated in the PARSEC platform, demonstrate notable improvements in responsiveness and efficiency, highlighting the potential for further advancements in serverless computing.