Characterizing Serverless Workloads in Production

The rise in popularity of serverless computing has led to increasing applications across diverse domains which generates a wide variety of workloads daily. An in-depth knowledge of various categories of workloads and their behaviours is instrumental. Workload characterization involves the process of analysing and grouping the workloads within an execution environment. To characterize serverless workloads, we propose a method based on classification and clustering approaches applicable in the production environment. The proposed method achieves an Adjusted Rand Index (ARI) of up to 96% for inter-function characterization and up to 97% for intra-function characterization in the production environment.