Skip to content

README.md

Latest commit

 

History

History
23 lines (14 loc) · 2.27 KB

README.md

File metadata and controls

23 lines (14 loc) · 2.27 KB
Raw

Section 5: Results data

This directory contains the experiment data used for all tables, plots, and statistical tests in the paper. It corresponds to Section 5 (Results and Analysis — data) of the paper.

Paper section: Section 5.

Contents

  • run_table.csv — Core artifact. One row per (run, repetition). Columns include: topology, system_size, throughput_rps, avg_latency_ms, p95_latency_ms, failure_rate, request_count, cpu_usage_avg, memory_usage_avg, energy, dram_energy, and derived metrics. All analysis in the paper is based on this file.

  • RUN_TABLE_COLUMNS_EXPLANATION.md — Describes every column in run_table.csv: meaning, data source (Locust / Prometheus / EnergiBridge), units, and how values are computed from the raw run folders.

  • raw_runs/ — Per-run, per-repetition folders (run_X_repetition_Y). Each folder contains:

    • Locust results: locust/ — locust_output.txt, results_stats.csv, results_stats_history.csv, results.html, results_failures.csv, results_exceptions.csv
    • Prometheus logs: prometheus_cpu.txt, prometheus_memory.txt — CPU and memory time series for the measurement window
    • EnergiBridge: energibridge_output.csv — RAPL energy (package + DRAM) samples
    • Metadata: measurement_start.txt, measurement_end.txt, namespace.txt

Provenance: Data produced by the Experiment Runner when running the experiment defined in 4_experiment_execution/ (Section 4). Run table and per-run directories correspond to 4_experiment_execution/experiment-runner/examples/mubench-benchmarking/experiments/mubench_phase3_benchmarking/.

Omitted file (GitHub size limit): One raw EnergiBridge file — raw_runs/run_15_repetition_5/energibridge_output.csv — exceeds GitHub’s 100 MB file limit and is not included in this repository. Its aggregated energy metrics are in run_table.csv. To obtain the raw file, re-run that run (run 15, repetition 5) using the instructions in 4_experiment_execution/HOW_TO_RUN_THE_EXPERIMENT.md.

Analysis: The Jupyter notebooks in 5_results_analysis/ (Section 5) use this data to produce the paper’s figures and tables. See 5_results_analysis/README.md.