⚡ FlashDB: Log-Structured Merge (LSM) Key-Value Store

A high-performance, persistent Key-Value store written in C++17, modeled after LevelDB and RocksDB. It uses an LSM-Tree architecture optimized for write-heavy workloads and includes a Bloom Filter to minimize unnecessary disk I/O.

🚀 Key Features

• MemTable: Implemented using a probabilistic Skip List for $O(\log N)$ inserts and lookups (replacing slow balanced trees).
• Persistence: Flushes in-memory data to binary SSTables (Sorted String Tables) on disk for crash recovery.
• Optimization: Integrated a custom Bloom Filter (3-hash) to reject non-existent keys in ~8 microseconds, bypassing disk reads.
• Performance: Capable of 1.5 Million writes/second (single-threaded).

🛠 Tech Stack

• Language: C++17
• Build System: CMake
• Techniques: Pointer arithmetic, Binary Serialization, Probabilistic Data Structures.

⚡ How to Run

mkdir build && cd build
cmake ..
make

# Run the Engine
./flashdb

# Recover/View Data from Disk
./sst_dump data_with_bloom.sst
👨‍💻 Author
Aarav Agnihotri - Built to explore database internals and storage engine architecture.