Distributed Key Value Store using Raft Consensus Algorithm with Leader Lease

This project presents a Distributed Key-Value Store implemented using the Raft Consensus Algorithm with Leader Lease. Leveraging the reliability and fault tolerance offered by Raft, the system ensures consistent data replication and fault recovery across a distributed network of nodes. By integrating leader lease functionality, it enhances the time complexity of various operations.

Built by - Mohammad Aflah Khan, Shivaansh Mital & Nishaant Rastogi

Getting Started

Prerequisites

To run this implementation, you need:

• Python 3.x
• ZeroMQ library (pyzmq)

Installation

1. Clone this repository to your local machine.
2. Install dependencies using pip:

   pip install -r requirements.txt
   

Usage

1. Modify the connections.json file to specify the network topology and node addresses.
2. Run the Raft nodes:

   python raft_node.py --nodeId <node_id> [--restarting <True/False>]
   

   • <node_id>: Unique identifier for the Raft node.
   • --restarting: Optional flag to indicate whether the node is restarting after a crash.

Once run correctly the nodes will automatically timeout elect leaders and work. You can then run client.py to send requests to the nodes, kill nodes and restart them and do all other fancy things.

Acknowledgments

• This implementation was developed as a course assignment in Distributed Systems course offered at IIITD.
• Resources consulted -
  • https://raft.github.io/
  • https://en.wikipedia.org/wiki/Raft_(algorithm)
  • https://thesecretlivesofdata.com/raft/
  • https://www.yugabyte.com/tech/raft-consensus-algorithm/