Serverless FaaS Scalability Evaluation: An ECG Signal Processing Use Case

Abstract

The serverless approach provides a completely new way of developing cloud services, service scalability, and elasticity by utilizing the container-level virtualization and abstraction. In this paper, we present a use case of an application that processes streaming electrocardiograms by implementing the Function as a Service approach. Several experiments are conducted to evaluate the scalability and elasticity performance by checking the following hypothesis: The system will be highly scalable keeping the same response time and throughput, for up to 7000 data streams. To check the validity of the hypothesis we will provide an experimental research on the following research questions to analyze the performance behaviour of response and throughput with the increased load of parallel data streams testing the following cases: A) Function that will generate a variable number of data streams, B) Functions that are invoked sequentially, and C) Functions that are invoked in parallel. Our use case proved that system scales horizontally and satisfies all requests. When the workload has linearly increased the system has linearly increased throughput and results in a similar response time for each individual request.