Week 4 Day 5: Mini Project - Mock Replicated KV Store

We will build a simulation of a 3-node database. We will write to all 3, and read from 2 to see how consistency works.

1. The Store (node.js)

Run 3 instances of this server on ports 3001, 3002, 3003.

const express = require('express');
const app = express();
app.use(express.json());

let data = {}; // In-memory DB

app.post('/write', (req, res) => {
    const { key, value, timestamp } = req.body;
    // Last Write Wins logic
    if (!data[key] || timestamp > data[key].timestamp) {
        data[key] = { value, timestamp };
    }
    res.json({ status: 'ok' });
});

app.get('/read/:key', (req, res) => {
    const { key } = req.params;
    res.json(data[key] || null);
});

const PORT = process.argv[2] || 3000;
app.listen(PORT, () => console.log(`Node running on ${PORT}`));

2. The Client (client.js)

This script acts as the Coordinator.

const axios = require('axios');

const NODES = ['http://localhost:3001', 'http://localhost:3002', 'http://localhost:3003'];

// Write Quorum (W=3)
async function write(key, value) {
    const timestamp = Date.now();
    const promises = NODES.map(node => 
        axios.post(`${node}/write`, { key, value, timestamp })
             .catch(e => null) // Ignore errors
    );
    await Promise.all(promises);
    console.log(`Wrote ${key}=${value} to all nodes.`);
}

// Read Quorum (R=2)
async function read(key) {
    const promises = NODES.map(node => 
        axios.get(`${node}/read/${key}`)
             .catch(e => null)
    );
    const results = await Promise.all(promises);
    
    // Find latest timestamp
    let latest = null;
    results.forEach(res => {
        if (res && res.data) {
            if (!latest || res.data.timestamp > latest.timestamp) {
                latest = res.data;
            }
        }
    });
    console.log(`Read ${key}:`, latest ? latest.value : 'Not found');
}

// Write then Read
write('name', 'Cody').then(() => read('name'));

Experiment

1. Start all 3 nodes. Run client. It works.
2. Kill Node 3003. Run client.
   • Write will succeed on 3001/3002.
   • Read will succeed (if it hits 3001/3002).
   • This proves Availability despite failure!

Next week: Messaging & Async. Decoupling systems with Kafka! 📨

Next Step

Next: Message Queues →