Skip to content

Benchmark Suite

Jump to bottom
Abdullah edited this page Feb 8, 2026 · 28 revisions

Benchmark Suite

The GraphBrew Benchmark Suite provides automated tools for running comprehensive experiments across multiple graphs, algorithms, and benchmarks.

Overview

scripts/
├── graphbrew_experiment.py     # ⭐ MAIN: One-click unified pipeline
│                                #    Downloads, builds, benchmarks, analyzes
├── requirements.txt            # Python dependencies
├── lib/                        # 📦 Core modules (all functionality)
│   ├── download.py             # Graph downloading
│   ├── benchmark.py            # Benchmark execution
│   ├── cache.py                # Cache simulation
│   ├── weights.py              # Weight management
│   ├── training.py             # ML training
│   ├── features.py             # Graph feature extraction
│   └── ...                     # Other modules
└── weights/                    # Auto-clustered type weights
    ├── active/                 # C++ reads from here
    ├── merged/                 # Accumulated weights
    └── runs/                   # Historical snapshots

🚀 Quick Start (One-Click Recommended)

The unified script handles everything automatically:

# Full pipeline: download → build → benchmark → analyze → train
python3 scripts/graphbrew_experiment.py --full --size small

# Just run benchmarks on existing graphs
python3 scripts/graphbrew_experiment.py --phase benchmark

# Quick test with key algorithms
python3 scripts/graphbrew_experiment.py --size small --quick

# Skip cache simulations (faster)
python3 scripts/graphbrew_experiment.py --phase all --skip-cache

# Run brute-force validation
python3 scripts/graphbrew_experiment.py --brute-force

# Complete training pipeline: reorder → benchmark → cache sim → update weights
python3 scripts/graphbrew_experiment.py --train --size small --max-graphs 5

Download Size Options

Size Graphs Total Size Use Case
SMALL 16 ~62 MB Quick testing
MEDIUM 28 ~1.1 GB Standard experiments
LARGE 37 ~25 GB Full evaluation
XLARGE 6 ~63 GB Massive-scale testing
ALL 87 ~89 GB Complete benchmark

Graph Categories

The graph catalog includes diverse graph types:

Category Examples Count
Mesh mesh networks, finite element 16
Web uk-2002, webbase-2001, web-Google 15
Collaboration hollywood-2009, com-DBLP, ca-AstroPh 11
Social soc-LiveJournal, com-Orkut 10
Synthetic RMAT, uniform random 8
Road roadNet-CA, europe-osm, USA-road-d 7
Citation cit-Patents, cit-HepPh 5
Commerce amazon0601, com-Amazon 5
P2P p2p-Gnutella24/25/30/31 4
Communication email-Enron, wiki-Talk 3
Infrastructure as-Skitter 3

All results are saved to ./results/:

  • reorder_*.json - Reordering times
  • benchmark_*.json - Execution times
  • cache_*.json - Cache miss rates (L1/L2/L3)
  • mappings/ - Pre-generated label maps for consistent reordering
  • graph_properties_cache.json - Cached graph properties for type detection

Auto-clustered weights are saved to ./scripts/weights/active/:

  • type_registry.json - Maps graph names to type IDs + cluster centroids
  • type_N.json - Per-cluster trained ML weights (type_0.json, type_1.json, etc.)

Alternative Approach (Manual)

If you prefer more control, you can use specific phases of the unified script:

1. Download Graphs

# List available graphs
python3 scripts/graphbrew_experiment.py --download-only --size small

# Download specific size category
python3 scripts/graphbrew_experiment.py --download-only --size medium

# Force re-download
python3 scripts/graphbrew_experiment.py --download-only --force-download

2. Generate Label Maps (Consistent Reordering)

# Pre-generate label maps for all algorithms
python3 scripts/graphbrew_experiment.py --generate-maps --size small

3. Run Specific Phase

# Just reordering phase (measure reorder times)
python3 scripts/graphbrew_experiment.py --phase reorder --size small

# Just benchmarks (skip cache simulation)
python3 scripts/graphbrew_experiment.py --phase benchmark --size small --skip-cache

# Just cache simulation
python3 scripts/graphbrew_experiment.py --phase cache --size small

# Just weight generation (from existing results)
python3 scripts/graphbrew_experiment.py --phase weights

4. Custom Configuration

# Specify graph size range
python3 scripts/graphbrew_experiment.py --min-mb 10 --max-mb 500

# Limit number of graphs
python3 scripts/graphbrew_experiment.py --max-graphs 10

# Custom trials
python3 scripts/graphbrew_experiment.py --trials 10

# Key algorithms only (faster)
python3 scripts/graphbrew_experiment.py --quick

Output Format

Benchmark Results (benchmark_*.json)

[
  {
    "graph": "email-Enron",
    "algorithm": "LeidenCSR",
    "algorithm_id": 17,
    "benchmark": "pr",
    "time_seconds": 0.0234,
    "reorder_time": 0.0,
    "trials": 2,
    "success": true,
    "error": "",
    "extra": {}
  }
]

Cache Results (cache_*.json)

[
  {
    "graph": "email-Enron",
    "algorithm_id": 17,
    "algorithm_name": "LeidenCSR",
    "benchmark": "pr",
    "l1_hit_rate": 85.2,
    "l2_hit_rate": 92.1,
    "l3_hit_rate": 98.5,
    "l1_misses": 0,
    "l2_misses": 0,
    "l3_misses": 0,
    "success": true,
    "error": ""
  }
]

Reorder Results (reorder_*.json)

[
  {
    "graph": "email-Enron",
    "algorithm_id": 17,
    "algorithm_name": "LeidenCSR",
    "time_seconds": 0.145,
    "mapping_file": "results/mappings/email-Enron/LeidenCSR.lo",
    "success": true,
    "error": ""
  }
]

Perceptron Weights (scripts/weights/active/type_N.json)

{
  "LeidenCSR": {
    "bias": 0.85,
    "w_modularity": 0.25,
    "w_log_nodes": 0.1,
    "w_log_edges": 0.1,
    "w_density": -0.05,
    "w_avg_degree": 0.15,
    "w_degree_variance": 0.15,
    "w_hub_concentration": 0.25,
    "cache_l1_impact": 0.1,
    "cache_l2_impact": 0.05,
    "cache_l3_impact": 0.02,
    "cache_dram_penalty": -0.1,
    "w_reorder_time": -0.0001,
    "_metadata": {
      "win_rate": 0.85,
      "avg_speedup": 2.34,
      "times_best": 42,
      "sample_count": 50,
      "avg_reorder_time": 1.234,
      "avg_l1_hit_rate": 85.2,
      "avg_l2_hit_rate": 92.1,
      "avg_l3_hit_rate": 98.5
    }
  }
}

PageRank Convergence Analysis

Analyze how reordering affects PageRank convergence.

Usage

Run PageRank directly via the binary with verbose output:

# Run PR with verbose convergence output
./bench/bin/pr -f graph.mtx -s -o 7 -n 5

Or include in the experiment pipeline:

# Run benchmarks (includes convergence data in results)
python3 scripts/graphbrew_experiment.py --phase benchmark --size small

Example Output

PageRank convergence varies by reordering algorithm:

Graph: facebook.el
┌────────────────────┬────────────┬──────────────┐
│ Algorithm          │ Iterations │ Final Error  │
├────────────────────┼────────────┼──────────────┤
│ ORIGINAL (0)       │ 18         │ 9.2e-7       │
│ HUBCLUSTERDBG (7)  │ 16         │ 8.8e-7       │
│ LeidenOrder (15)   │ 15         │ 9.1e-7       │
│ LeidenCSR (17)     │ 14         │ 8.5e-7       │
└────────────────────┴────────────┴──────────────┘

Experiment Workflow

Complete Reproducible Experiment (One Command)

# One-click full experiment
python3 scripts/graphbrew_experiment.py --full --size medium

This automatically:

  1. Downloads graphs from SuiteSparse
  2. Builds binaries
  3. Generates label maps for consistent reordering
  4. Runs all benchmarks with all algorithms
  5. Runs cache simulations
  6. Trains perceptron weights with cache + reorder time features

Custom Experiment Workflow

#!/bin/bash
# Custom experiment workflow

cd GraphBrew

# 1. Download specific size graphs
python3 scripts/graphbrew_experiment.py --download-only --size medium

# 2. Generate label maps once
python3 scripts/graphbrew_experiment.py --generate-maps --size medium

# 3. Run benchmarks using pre-generated maps
python3 scripts/graphbrew_experiment.py --use-maps --phase benchmark --size medium --trials 10

# 4. Run cache simulations
python3 scripts/graphbrew_experiment.py --use-maps --phase cache --size medium

# 5. Generate perceptron weights
python3 scripts/graphbrew_experiment.py --phase weights

# 6. Run brute-force validation
python3 scripts/graphbrew_experiment.py --brute-force --size medium

Parallel Execution

Thread Control

# Control OpenMP threads
export OMP_NUM_THREADS=8
python3 scripts/graphbrew_experiment.py --full --size small

Skip Heavy Operations

# Skip slow algorithms on large graphs
python3 scripts/graphbrew_experiment.py --skip-slow --size large

# Skip expensive cache simulations (BC, SSSP)
python3 scripts/graphbrew_experiment.py --skip-expensive --phase cache

Troubleshooting

"Graph not found"

# Check downloaded graphs
ls results/graphs/

# Re-download with force flag
python3 scripts/graphbrew_experiment.py --download-only --force-download

Timeout Issues

# Increase timeouts
python3 scripts/graphbrew_experiment.py \
    --timeout-reorder 86400 \
    --timeout-benchmark 7200 \
    --timeout-sim 14400

Memory Issues

# Use smaller graphs
python3 scripts/graphbrew_experiment.py --size small

# Skip large graphs
python3 scripts/graphbrew_experiment.py --max-mb 500

Clean Start

# Clean results only (keep graphs)
python3 scripts/graphbrew_experiment.py --clean

# Full reset (remove everything including downloaded graphs)
python3 scripts/graphbrew_experiment.py --clean-all

Next Steps

← Back to Home | Correlation Analysis →

Clone this wiki locally