Single Node Benchmark Workflow

System benchmarking workflow for the Parallel Works ACTIVATE platform. Runs CPU, memory, and disk I/O benchmarks on a single node and displays results through an interactive Plotly-based web visualization.

Quick Start

Run Locally

source ~/venv/bin/activate
uv pip install -r requirements.txt
python scripts/local_runner.py --duration 10

Deploy to ACTIVATE

1. Push this repository to your ACTIVATE account
2. Select the workflow from your workflows list
3. Configure benchmark options and target cluster
4. Run and view interactive results via the tunnel session

Workflow Inputs

Input	Type	Default	Description
cluster	compute-clusters	-	Target cluster for benchmark execution
duration	number	10	Duration for each benchmark in seconds (5-60)
run_cpu	boolean	true	Run CPU benchmark
run_memory	boolean	true	Run memory benchmark
run_disk	boolean	true	Run disk I/O benchmark

Benchmarks

Benchmark	Metric	Description
CPU	ops/sec	Prime number calculations measuring single-thread performance
Memory	MB/s	Sequential memory read/write throughput
Disk I/O	MB/s	Sequential file read/write using 256MB test file

CPU Benchmark

Calculates prime numbers using trial division. Measures operations per second and tracks the number of primes found during the test duration.

Memory Benchmark

Allocates a 100MB buffer and performs sequential write and read operations. Measures throughput in MB/s for both operations.

Disk I/O Benchmark

Uses dd to write a 256MB test file and then read it back. Measures sequential throughput in MB/s. The test file is automatically cleaned up after the benchmark.

Results

Benchmark results are saved as:

• benchmark_results.json - Raw results in JSON format
• benchmark_results.html - Interactive Plotly visualization

The HTML visualization includes:

• Bar charts for each benchmark type
• System information table
• Hover tooltips with detailed metrics

Development

Setup

# Activate shared virtual environment and install development dependencies
source ~/venv/bin/activate
uv pip install -r requirements-dev.txt

Run Tests

# Run all tests
pytest

# Run with coverage
pytest --cov=scripts --cov-report=html

# Skip slow integration tests
pytest -m "not slow"

Local Runner Options

# Quick test with short duration
python scripts/local_runner.py --duration 3

# Run only CPU benchmark
python scripts/local_runner.py --cpu-only --duration 5

# Run and serve results in browser
python scripts/local_runner.py --duration 10 --serve

# Specify custom output directory
python scripts/local_runner.py --output-dir ./my_results --serve --port 9000

Project Structure

activate-benchmark/
├── workflow.yaml           # ACTIVATE workflow definition
├── scripts/
│   ├── __init__.py         # Package marker
│   ├── run_benchmarks.sh   # Benchmark execution script
│   ├── run_benchmarks.py   # Python module for benchmarks (testable)
│   ├── generate_plot.py    # Interactive plot generation (Plotly)
│   ├── serve_results.py    # Simple HTTP server for results
│   └── local_runner.py     # CLI tool to run workflow locally
├── tests/
│   ├── __init__.py
│   ├── conftest.py         # Pytest fixtures
│   ├── test_benchmarks.py  # Unit tests for benchmark functions
│   ├── test_plot.py        # Unit tests for plot generation
│   └── test_integration.py # Integration tests for full workflow
├── results/                # Output directory (created at runtime)
├── requirements.txt        # Runtime dependencies
├── requirements-dev.txt    # Development/test dependencies
├── pytest.ini              # Pytest configuration
└── README.md               # This file

Dependencies

• Python 3.8+
• uv (fast Python package installer)
• plotly (Python package for visualization)
• Standard system tools: dd, python3
• Shared virtual environment at ~/venv

Troubleshooting

Disk benchmark shows 0 MB/s

The disk benchmark requires write access to the temp directory. Ensure sufficient disk space is available.

Memory benchmark is slow

The memory benchmark allocates 100MB of memory. On systems with limited RAM, this may cause swapping.

Cannot access results via browser

Ensure the ACTIVATE tunnel session is properly configured and your browser allows popups from the ACTIVATE domain.