High-Performance CPU Scheduling Benchmark Suite

A professional C++17 benchmarking suite for OS scheduling algorithms. Designed for high-concurrency evaluation and architectural performance analysis.

Architecture

The system follows a producer-consumer model for multi-core simulation.

graph TD
    A[Workload Generator] --> B[Scheduler Engine]
    B --> C{Ready Queue}
    C -->|Mutex| D[ThreadPool Workers]
    C -->|Lock-Free| D
    D --> E[Metrics Collector]
    E --> F[CSV Export]
    E --> G[Visualization Script]

Loading

Supported Algorithms

1. FCFS: First-Come, First-Served.
2. SJF: Shortest Job First.
3. Round Robin: Time-sliced preemptive scheduling.
4. Priority: Priority-based selection.

Concurrency Model

The core of the benchmark is the Michael-Scott Lock-Free Queue, which uses atomic compare_exchange operations to minimize synchronization overhead in high-concurrency scenarios.

Build Instructions

Linux (Make)

make release
make test

Windows (MSVC)

./build.bat
./build.bat test

Usage Examples

Single Algorithm Benchmark

./scheduler --algo RR --processes 10000 --cores 8 --queue lockfree

Full Stress Test

./scheduler --stress-test --cores 16 --queue lockfree

Automated Benchmarking

Run the full experiment sweep and generate visualizations:

# Windows
./benchmarks/run_benchmark.bat
python scripts/plot_results.py

# Linux
./benchmarks/run_benchmark.sh
python scripts/plot_results.py

Benchmark Results

The following results were obtained on an 8-core simulated environment with 10,000 processes.

Performance Visualizations

Performance Profiling

Analyze the scheduler efficiency using linux-perf:

# General stats
perf stat ./scheduler --processes 100000 --cores 16

# Cache and Branch analysis
perf record -e cache-misses,branch-misses ./scheduler
perf report

---

Developed as a high-performance systems engineering benchmark suite.