Network Based Performance Analysis of Microservice Behavior in Containerized and Non-Containerized Environment

Abstract

In contemporary software design, microservices play a central role in facilitating scalable, fault-tolerance, and autonomous service deployments. Deploying microservices in containerized and non-containerized environments, however, poses serious issues about network performance, latency, and resource usage. This study solves the issue of knowing the tradeoffs in microservice deployment between these two environments. The main objective is to assess and compare network-based performance metrics like latency, throughput, and CPU usage. To achieve this, an experimental methodology was adopted whereby identical microservices were run in Docker-based (containerized) and VM-based (noncontainerized) setups. Prometheus and Grafana were employed for real-time monitoring, and Locust for simulating Application Programming Interface(API) load. The findings indicated that containerized environments provided significant savings in Central Processing Unit usage by as much as 25%, and better response times at medium to high traffic levels. On the other hand, non-containerized setups provided lower network latency by around 18%, pointing to a compromise between communication speed and system efficiency. In this research, it is found that although containerization maximizes resource utilization and scalability, there is a little network overhead introduced. Finally, this work provides support for data-driven decisions for system architects when deploying cloud-native applications.

Topics & Concepts