Horizontal vs Vertical Scaling

Why This Matters

Your application works great with 100 users. Then it hits 10,000 users and slows to a crawl. Then 100,000 users arrive and the server crashes. How do you handle more load? You have two options: vertical scaling (get a bigger machine) or horizontal scaling (add more machines). This decision shapes your architecture, your costs, and your ability to grow.

Vertical scaling is straightforward -- upgrade CPU, RAM, or storage. But every machine has a ceiling, and a single machine is a single point of failure. Horizontal scaling adds more servers behind a load balancer, but now you must handle shared state, session management, and data consistency across machines. Most modern systems use a combination of both strategies. Understanding when and how to apply each approach is fundamental to building systems that grow with your users.

Define Terms

Visual Model

Code Example

// Simulating a load balancer with round-robin
class LoadBalancer {
  constructor(servers) {
    this.servers = servers;
    this.current = 0;
  }

  getNextServer() {
    const server = this.servers[this.current];
    this.current = (this.current + 1) % this.servers.length;
    return server;
  }

  handleRequest(request) {
    const server = this.getNextServer();
    console.log(`Routing to ${server}: ${request}`);
    return server;
  }
}

// Horizontal scaling: 3 servers behind a load balancer
const lb = new LoadBalancer([
  "server-1:3000",
  "server-2:3000",
  "server-3:3000",
]);

// Simulate 6 requests — distributed evenly
for (let i = 1; i <= 6; i++) {
  lb.handleRequest(`Request ${i}`);
}
// Output:
// Routing to server-1:3000: Request 1
// Routing to server-2:3000: Request 2
// Routing to server-3:3000: Request 3
// Routing to server-1:3000: Request 4
// ...

Interactive Experiment

Try these exercises:

• Modify the load balancer to use "least connections" instead of round-robin. Track how many active requests each server has.
• Simulate a server going down: remove a server from the list. How does the load balancer adapt?
• Think about session state: if a user logs in on server-1 and the next request goes to server-2, what happens? How would you solve this?
• Calculate the cost difference: one server with 64 GB RAM vs four servers with 16 GB RAM each. Which is more cost-effective?

Quick Quiz

Coding Challenge

Real-World Usage

Scaling strategies are essential in production:

• AWS Auto Scaling: Automatically adds or removes EC2 instances based on CPU utilization, network traffic, or custom metrics.
• Kubernetes Horizontal Pod Autoscaler: Scales the number of pods in a deployment based on observed CPU or memory usage.
• Database read replicas: Databases like PostgreSQL and MySQL scale reads horizontally by replicating data to read-only copies.
• Vertical scaling for databases: It is common to vertically scale database servers (more RAM, faster SSDs) because horizontal scaling of stateful systems is much harder.
• CDNs: Content delivery networks scale horizontally by adding edge servers worldwide to serve cached content closer to users.

Connections