Measuring Function-as-a-Service Performance in Different Cloud Platforms using an Application-Level Benchmark

Abstract This paper extends earlier work assessing the performance of serverless computing platforms, or Function as a Service (FaaS), across major public cloud providers, Google Cloud Platform (GCP), Azure, and Amazon Web Services (AWS), using an application-level benchmark. It aims to reproduce previous results, observe possible performance changes over time, and improve the experimental set-up for a fairer comparison of the cloud service providers. The benchmark application simulates a common user account creation process, including both synchronous and asynchronous tasks. By collecting timestamps from different stages of the workflow, the study measures response times, execution durations, and variation across runs. The findings show that function execution is generally consistent across platforms, but the triggering of functions is more variable. Azure functions, in particular, demonstrate longer and less stable trigger times compared to AWS and GCP. By contrast, AWS and GCP provide faster and more predictable triggers, though with some differences in asynchronous behavior. The results confirm the earlier study’s conclusions to a large extent, while also offering new insights thanks to the adjustments in methodology. These findings are relevant for organizations that need to decide which platform to use for serverless applications and also contribute to longer-term benchmarking of FaaS performance. Future improvements could include extending the benchmark to additional workflows and platforms, as well as increasing automation in deployment and data collection.