Evaluating scalability and performance in microservices architectures
Dung, Ho (2026)
2026
Master's Programme in Computing Sciences and Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2026-05-05
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202605044904
https://urn.fi/URN:NBN:fi:tuni-202605044904
Tiivistelmä
Microservices architecture has become the main approach for building scalable, distributed software systems. However, the choice of inter-service communication protocols and container orchestration platforms appears to have a clear effect on system performance and scalability, and there is not enough empirical evidence to guide these decisions in practice.
This thesis presents an empirical evaluation of scalability and performance in microservices architectures, with a focus on two communication protocols, REST and gRPC, and two container orchestration platforms, Docker Swarm and Kubernetes. The study follows a quantitative experimental approach: identical microservice applications are deployed under controlled conditions, and key performance indicators, including response time, throughput, error rate, CPU utilization, and memory consumption, are measured across varying workloads.
Performance data is collected using an integrated monitoring stack that includes Prometheus for metrics aggregation, Jaeger for distributed tracing, and Grafana for visualization. Load testing is carried out using Grafana K6 to simulate realistic traffic patterns at multiple concurrency levels.
The results indicated that gRPC outperformed REST in the tested scenarios in terms of response time and throughput, particularly under high-load conditions, owing to its binary serialization through Protocol Buffers and HTTP/2 multiplexing. Kubernetes was observed to demonstrate stronger auto-scaling capabilities and fault tolerance compared to Docker Swarm, maintaining stable performance under increasing workloads. The integrated monitoring stack was found to be effective for identifying performance bottlenecks and supporting data-driven optimization decisions.
These findings provide practitioners with evidence-based guidance for choosing communication protocols and orchestration platforms in microservices deployments. The study addresses a notable gap in the empirical literature on microservices performance engineering.
This thesis presents an empirical evaluation of scalability and performance in microservices architectures, with a focus on two communication protocols, REST and gRPC, and two container orchestration platforms, Docker Swarm and Kubernetes. The study follows a quantitative experimental approach: identical microservice applications are deployed under controlled conditions, and key performance indicators, including response time, throughput, error rate, CPU utilization, and memory consumption, are measured across varying workloads.
Performance data is collected using an integrated monitoring stack that includes Prometheus for metrics aggregation, Jaeger for distributed tracing, and Grafana for visualization. Load testing is carried out using Grafana K6 to simulate realistic traffic patterns at multiple concurrency levels.
The results indicated that gRPC outperformed REST in the tested scenarios in terms of response time and throughput, particularly under high-load conditions, owing to its binary serialization through Protocol Buffers and HTTP/2 multiplexing. Kubernetes was observed to demonstrate stronger auto-scaling capabilities and fault tolerance compared to Docker Swarm, maintaining stable performance under increasing workloads. The integrated monitoring stack was found to be effective for identifying performance bottlenecks and supporting data-driven optimization decisions.
These findings provide practitioners with evidence-based guidance for choosing communication protocols and orchestration platforms in microservices deployments. The study addresses a notable gap in the empirical literature on microservices performance engineering.