Usage

Prerequisites

• Cluster with Kepler and Prometheus deployed
  • Prometheus must scrape cAdvisor metrics (tested with kube-prometheus-stack)
• kubectl is correctly configured to manage your target cluster
• Kepler metrics are exported to Promtheus server
• Prometheus server is available at http://localhost:9090. Otherwise, set environment variable: PROM_SERVER

Run scalability benchmarks

Run ./script.sh run_benchmark

Plot overhead metrics from benchmark files

1. Open Jupyter Notebook
2. Assign result_directories and experiment_names values
   • result_directories should be the paths to the directories created by ./script.sh run_benchmark (e.g. _output/results/scale-dummy-deployment/2023_10_04_09_47_AM)
   • experiment_names will be the labels used when plotting multiple experiments and should be aligned to match the indexing of result_directories
3. Execute plot_experiments(result_directories, experiment_names)