OS3: A Step Forward Towards Enhancing Share and Reuse in Serverless Functions
Sarvesh Bhatnagar · Deep Blue (University of Michigan) · 2023
Known for its execution flexibility and pricing elasticity, Serverless computing deploys the code of services on demand and charges the client for their actual execution. Serverless computing enables service developers to focus on creating useful services, without being concerned about how these services would be deployed and provisioned. These developments have fostered the rapid growth of a community of open-source developers creating and improving various serverless functionalities. Because of this variety, the ecosystem of serverless functionalities suffers from a high degree of duplication, with multiple serverless functionalities sharing similarities or being outright identical. In order to integrate a serverless function into a project, a developer first needs to be able to find a suitable implementation among multiple alternatives. Unfortunately, existing search technologies have not been designed to address the needs of searching for serverless functionalities. To address this problem, this paper presents a novel approach to search for serverless functions, called Open-Source Serverless Search (OS3) that maximizes the utility of the returned serverless functions by (1) basing the search process on both descriptive keywords and library usages, thus increasing the search results' precision and completeness; (2) filtering and ranking the search results on both the software license and execution cost parameters. Implemented in 3,000 lines of Python, with a search space of 5,981 serverless repositories from four major serverless platforms, OS3 outperforms existing search approaches in terms of the suitability of the returned serverless functions, based on our evaluation with realistic use cases. Enhancing the search facilities for serverless functions with the insights presented herein can help fully fulfill the enormous promise of serverless computing.