Raft Consensus Algorithm

Overview

Raft is a consensus algorithm for managing the replicated log over a cluster of servers in a consistent state. This project is an implementation of the Replicated Key-Value Store on top of the Raft Consensus Protocol. This repository contains the assignments for the course "CS-733 Engineering the Cloud- Advanced Distributed Computing" offered at IIT Bombay.

Project Structure

The project at its core has two parallels, one is the Raft side and other is the Key-Value Store. These parallels run as Go-routines and communicate among themselves via channel. The complete description of the Raft package can be found here.

Execution Instructions

To build the project use following command.

go get github.com/pankajrandhe/cs733/assignment4/

To test the project execute the following command.

go test github.com/pankajrandhe/cs733/assignment4 -v

Usage

Following are the commands which Key-Value store accepts from the client.

• SET command

  set <key> <exptime> <numbytes> [noreply]\r\n

  <value bytes>\r\n

  Server responds with:

  OK <version>\r\n

• GET command

  get <key>\r\n

  Server responds with:

  VALUE <numbytes>\r\n

  <value bytes>\r\n

• GETM command

  getm <key>\r\n

  Server responds with:

  VALUE <version> <exptime> <numbytes>\r\n

  <value bytes>\r\n

• CAS command

  cas <key> <exptime> <version> <numbytes> [noreply]\r\n

  <value bytes>\r\n

  Server responds with:

  OK <version>\r\n

• DELETE command

  delete <key>\r\n

  Server responds with:

  DELETED\r\n

BUGS

• The project has not yet been tested for the log-reapair test cases.

References

• In Search of an Understandable Consensus Algorithm by Diego Ongaro and John Ousterhout, Stanford University.
• Generating random number in a given range
• File Append Operations
• Ticker Implementation
• Discussions with Prof. Sriram Srinivasan
• Help from friends: Abhishek Potnis, Bhavesh Singh, Dushyant Sabharwal, Saurabh Hote and discussions on Piazza forum for CS733.