Raft-Implementation

About the project

• This repository contains a simple RAFT consensus algorithm with leader lease implemented as part of the CSE530 course offered at IIIT-D.
• Use the files node.py and client.py to run the RAFT algorithm locally.
• Use the files gcp-node.py and gcp-client.py to run the RAFT algorithm on Google Cloud Platform (or any other cloud instance).
• Team Members: Jaskaran Singh, Gautam Reddy

Introduction

• Raft is a consensus algorithm designed to be easy to understand. Raft divides time into terms, and each term begins with an election to select a leader. The leader manages the replication of the log entries to the other servers and ensures a high degree of consistency.
• This implementation of RAFT also employs leader lease for optimized read only queries. Leases can be considered as tokens that are valid for a certain period of time, known as lease duration. A node can serve read and write requests from the client if and only if the node has this valid token/lease.
• For a more detailed explanation please refer the following page.

Remote Procedure Calls (RPCs)

• Raft servers communicate using remote procedure calls (RPCs), and the basic consensus algorithm requires only two types of RPCs.
• RequestVote RPCs are initiated by candidates during elections.
• AppendEntries RPCs are initiated by leaders to replicate log entries and to provide a form of heartbeat.
• For this project, gRPC has been used to facilitate RPC functionality.
• The RPCs utilized in this project are detailed in the below figure (Note: The RPCs contain slight modifications to incorporate the leader lease functionality):

References

• RAFT explanation articles: PART-1, PART-2
• Low Latency reads using Leader Lease: Yugabyte Blog

Acknowledgments

• Election animation GIF source: Alexander Brooks's Medium article
• README.md: Link