Analysing the Effects of Scalability in Microservices to User-Perceived Performance and Cloud Costs

Abstract

Microservices are widely adopted in cloud computing systems, where scalability is essential for maintaining performance and controlling operational costs under rapidly changing workloads. While cloud computing platforms provide elastic infrastructure, the ability of microservices to benefit from elasticity depends on their scalability. This thesis analyses how various levels of scalability in microservices affect user-perceived performance and cloud costs, and identifies operational characteristics that drive scalability.

This thesis adopts an empirical and quantitative methodology combining a conceptual analysis of microservice implementation technologies, a structured literature review and controlled empirical experiments. Microservice benchmark applications with various levels of scalability are compared under simulated workloads to compare latency and resource usage. Java-based microservice frameworks are used to compare the implications of scalability in a common runtime ecosystem, while the analysis remains framework and language agnostic.

The results show that scalability in microservices is driven by application startup time, container image size, resource efficiency and request throughput. Poor scalability manifests as increased tail-latency, latency spikes during scale-out and unpredictable response times, thus degrading overall user experience. In contrast, improved scalability enables high resource-efficiency, reduces the need for resource over-provisioning and thus leads to lower cloud computing costs.