Evaluation of scalability in the Fission serverless framework

Abstract

. The efficient code execution often requires concurrency, so many programming languages, libraries and framework aim at parallelism. Based on the granularity and abstraction level, many approaches of concurrency are available. However, concurrency carries difficulties but modern ways try to make it more convenient. A rather new solution is cloud computing which enhances the concurrency in a way that standalone virtual machines utilize the shared hardware. Con-tainerization takes advantage of lightweight virtual machines called containers because they use a shared kernel of the operating system. Conteiner orchestration (e.g. Kubernetes) enables containerization among multiple hosts. Serverless programming supports container orchestration for individual function so every trigerred function may run in a different container which is inside a cluster of hosts. In this paper, we briefly present the modern cloud computing ways of concurrency. This subtle distributed approach requires a comprehensive evaluation. We take advantage of the open source Fission serverless framework and implement some distributed algorithms in this realm using the Python programming language. For a deeper comprehension, we measure and evaluate the scalability of Fission framework and the entire system. We execute the distributed algorithms with different sizes of input and we fine-tune the configuration of the Fission framework.