Monitoring Pace Scheduler

Overview

The Monitoring Pace Scheduler is an adaptive monitoring solution that dynamically adjusts scrape intervals based on metric patterns. Unlike traditional fixed-interval approaches, it analyzes metric change frequencies to automatically optimize scrape intervals, reducing data collection during stable periods and increasing frequency when metrics show high activity.

Prerequisites

Clone the repository:

git clone https://gitlab.com/your-repo/monitoring-pace-scheduler.git
cd monitoring-pace-scheduler

Install dependencies:

pip install -r requirements.txt

Install tcpdump:

sudo apt install tcpdump

Additional components:

• Prometheus & Node Exporter – Setup Guide
• (Optional) Gatling for load simulation – Simulation Setup

---

Configuration

Set up the Monitoring Behavior

Edit config.yaml to define scrape logic:

prometheus:
  config_file: 'prometheus.yml'             # Path to the Prometheus configuration file
  reload_url: 'http://0.0.0.0:9091/-/reload' # URL to trigger Prometheus config reload

thresholds:
  update_threshold: 0.05       # Minimum change ratio required to update the scrape interval
  default_scrape_interval: 15  # Initial interval (in seconds)
  max_scrape_interval: 900     # Maximum allowed scrape interval (in seconds)

metrics:
  to_monitor:
    - '(1 - avg(rate(node_cpu_seconds_total{mode="idle"}[1m])) by (instance)) * 100'  # metric to monitor : CPU usage percentage

---

Execution

To run the adaptive monitoring strategy in a production-like environment:

# Run adaptive group 
python3 scheduler.py --duration 3600

By default, the scheduler runs for 1 hour (--duration 3600)

You can remove the --duration flag to run the monitoring loop continuously.

---

Evaluation

To assess the impact and efficiency of dynamic scraping:

Benchmark Steps:

• Use Gatling to simulate fluctuating system loads

• Capture Prometheus network traffic using tcpdump

• Compare baseline vs. dynamic strategies in terms of:

  • Data volume sent
  • Metric behaviour

Full procedure: Benchmark Instructions

Evaluation notebook: src/benchmark/evaluation.ipynb

---

License

This project is licensed under the MIT License.