An Empirical Study on Impact of Programming Languages on Performance of Open-source Serverless Platforms

The concept of serverless computing is an emerging offering in the cloud landscape that promotes a higher level of abstraction and further separates software operations from the hardware platform. Using serverless computing, stateless functions are executed in a short period of time, enabling finer-grained control and accounting of resources. Open-source serverless computing frameworks provide an attractive alternative to public cloud platforms and also offer the possibilities of serverless computing for on-premise deployments. This paper examines three popular open-source serverless platforms that run on a Kubernetes cluster: OpenFaaS, Nuclio and Fission. We used CloudLab, a scientific infrastructure for research on the next generation of computing platforms, as our test environment. Different functions are defined using two popular programming languages, Python and Node.js, and invoked at different concurrency levels. For this purpose, Hey and Wrk are used to generate the required workloads. After that, the performance of the mentioned platforms when invoking the defined functions is evaluated and analyzed in terms of response time, throughput, and total data exchanged. The results of the experimental evaluation demonstrate that if OpenFaaS is the platform of choice, Python is a better choice as the programming language than Node.js. If Fission is chosen, Node.js outperforms Python, especially at higher concurrencies. If we choose Nuclio, none of the considered programming languages is the sole winner. Regardless of the programming language, Nuclio performs better than other platforms.