fastDB

fastDB is an in-memory Bank database program written in low-latency C++ with multi-threading with a Python backend.It uses a custom multi threaded architecture to process transactions asynchronously and maintain real-time indexes along with user balances

Overview

fastDB uses custom data structures(skip lists,segment trees,Thread safe queue,heaps) to minimize latency for continuous querying and database updation and insertion.

• Engine: C++ for low-latency data insertion and querying
• Bindings: pybind11 Python module to bind c++ code and use it as a library inside python
• API: FastAPI server for handling web traffic and API
• Client: HTML/CSS/JS dashboard

---

Performance

fastDB uses advanced data structures(skip lists and segment trees and heap) to achieve $O(1)$ and $O(\log N)$ time complexities for core operations.

Operation / Endpoint	Time Complexity	Underlying Data Structure
Register User (/register)	$O(\log N)$	Skip List + Segment Tree
Transfer Money (/transfer)	$O(1)$ API $O(\log N)$ Engine	Thread-Safe Queue (API) + Skip List + Segment Trees
Check Balance (/balance/{id})	$O(1)$	Direct Array Indexing (std::vector)
Leaderboard (/leaderboard)	$O(K)$	Min/Max Heap ($K$ is top_n)
Regional Wealth (/balanceinrange)	$O(\log N)$	Min/Max or Sum Segment Trees

Architecture

fastDB uses a single C++ worker thread that runs continuously,isolated from the main web server thread.The Python API pushes transaction requests into a queue,and mutex locks ensure the web server and database engine don't collide.

• Direct Memory Arrays(std::vector):User balances map directly to array indices.The engine accesses balances at balances[userid] for $O(1)$ read/write operations.
• Thread-Safe Queues (std::queue + std::mutex):Transfer requests are pushed to a C++ queue in $O(1)$ time.A background thread processes the queue asynchronously.
• Skip List (std::set):Balances are stored in nodes in skip list(multiple linked list )data structure.On a transaction,the old balance is removed and the new one inserted in $O(\log N)$ time to keep global rankings sorted for leaderboard queries.
• Segment Trees:The engine uses three concurrent Segment Trees for wealth queries in a range (Sum, Min, Max).These update on every transaction $O(\log N)$ time complexity.

---

Tech Stack

• Core Engine: C++11
• Python Bindings: pybind11
• Backend API: Python 3.8+,FastAPI,Uvicorn
• Frontend UI: HTML5,CSS3,Vanilla JavaScript

---

Prerequisites

• A C++ compiler
• Python 3.8+

pip install fastapi uvicorn pybind11

Commands to run

• step-1 c++ -O3 -Wall -shared -std=c++11 -fPIC$(python3 -m pybind11 --includes) main.cpp -o fastdb$(python3-config--extension-suffix)
• step-2 uvicorn main:app --reload
• step-3open in a new terminal and run = python3 -m http.server 8080