ParallelExperiment

A Java-based performance experiment that measures the speedup of parallel computation across different file sizes using multiprocessing and Bubble Sort.

🚀 Features

📁 File Generation

• Automatically generates 3 datasets of 120 files each
• Small files — 2000 numbers/file
• Medium files — 5000 numbers/file
• Large files — 10000 numbers/file
• Uses a fixed random seed to ensure reproducible results every run

⚙️ Parallel Processing

• Uses ProcessBuilder to spawn real OS-level child processes
• Dynamically splits 120 files across P workers (P = 1, 2, 3... up to CPU core count)
• Measures makespan — total time from first process start to last process finish

🔢 Bubble Sort O(n²)

• Each worker reads a file and sorts it using Bubble Sort
• O(n²) complexity ensures computation is heavy enough to show real parallelism gains
• Results are intentionally discarded — only timing matters