NANOSTORE: High-Performance C++ Redis-like Database

Description

Nanostore is a key-value database supporting three core operations: SET, GET, and DEL. It achieves high-speed request processing by utilizing a thread pool architecture with blocking queues and database partitioning. While designed natively for Linux, it can be run on Windows via WSL.

Usage

You can run the binaries located inside the bin/ folder:

• Nanostore The core database binary. This version runs entirely in memory and does not retain data after a crash (Non-persistent).

• Nanostore_persistent The persistent version of the database. It retains data after a crash but is slower due to disk I/O overhead (uses std::flush).

• client A CLI tool used to communicate with the database.

• stress_test A benchmarking tool that tests the database for a fixed number of requests.

• stress_duration A benchmarking tool that tests the database for a fixed time interval (default: 60s). Note: To change the duration, modify the value in the test/ folder and recompile.

Benchmarks

System Specs: Ryzen 5 5500U, 16GB RAM, WSL2 on Windows.

Version	Test Type	Throughput
Nanostore	Stress_test	961,606.67 req/sec
Nanostore	Stress_duration	2,626,585.00 req/sec
Nanostore_persistent	Stress_test	14,596.40 req/sec
Nanostore_persistent	Stress_duration	31,721.70 req/sec

Future Updates

• Optimize Persistence: Improve file writing speed (currently slow due to direct synchronous command writing).
• Error Handling: Add distinct error types in Types.hpp for better modularity and management.
• EPOLL Implementation: Migrate to epoll for event notification without degrading performance.