Z-ENG Performance and Scalability Evaluation of Kubernetes Container Networking Architectures

2025-2026 tavasz

Nincs megadva

Téma leírása

The widespread adoption of cloud-native architectures has made Kubernetes the dominant platform for deploying and managing containerized applications. In such environments, efficient container networking is essential for the performance and scalability of microservice-based systems. Kubernetes relies on Container Network Interface (CNI) plugins to implement networking between pods and services, but different CNI solutions follow different architectural approaches, such as overlay networking, routing-based networking, and eBPF-based dataplanes. These architectural differences can significantly affect network latency, throughput, resource consumption, and scalability in distributed systems. This thesis investigates the performance characteristics of representative Kubernetes networking architectures through a controlled and reproducible benchmarking study. The evaluation will compare commonly used CNI solutions including Flannel (overlay-based), Calico (routing-based), and Cilium (eBPF-based). Experiments will be conducted in a multi-node Kubernetes cluster where hardware configuration, Kubernetes version, and workload parameters are strictly controlled. The study will measure metrics such as TCP throughput, UDP packet loss and jitter, round-trip latency (RTT), and CPU and memory overhead. Experiments will include different communication scenarios (eg, same-node and cross-node pod communication, and service-based communication) as well as scalability tests with increasing numbers of pods and nodes. The data collected will be analyzed to identify performance differences and trade-offs between the evaluated networking architectures.

 

Tasks to be performed by the student will include:

  • Review related work on Kubernetes networking and CNI plugins.
  • Design a reproducible benchmarking methodology and experimental setup.
  • Deploy and configure a multi-node Kubernetes cluster.
  • Install and configure selected CNI plugins (eg, Flannel, Calico, and Cilium).
  • Implement and automate benchmark tests measuring throughput, latency, jitter, and packet loss.
  • Measure CPU and memory overhead during experiments.
  • Conduct scalability experiments with increasing numbers of pods and nodes.
  • Analyze experimental results and document the findings in a scientific thesis

Maximális létszám: 5 fő