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

Open Jupyter Notebook
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
Execute plot_experiments(result_directories, experiment_names)