Course contentsShow
← System Design
Lesson 1708 of 1,91941. Real-World System: Notification SystemFree lesson

Scalability and Horizontal Expansion

Scaling workers horizontally, partitioning queues, and handling increased load without degrading latency.

Scalability and Horizontal Expansion

What You'll Learn

You'll learn how to scale notification processing workers horizontally and partition queues to handle massive load spikes without degrading delivery latency.

The Challenge of Growing Load

As your notification system grows from thousands to millions of users, a single worker processing messages from a queue becomes a bottleneck. If Black Friday triggers 10 million notifications in an hour, one worker can't keep up—messages pile up, latency skyrockets, and users experience delays.

Horizontal Scaling Strategy

Horizontal expansion means adding more worker instances rather than making one worker bigger (vertical scaling). Think of it like opening multiple checkout lanes at a grocery store instead of trying to make one cashier work faster.

Key Techniques

Worker Parallelization: Deploy multiple identical worker instances that all pull from the same queue. Each worker independently processes messages, multiplying your throughput. If one worker handles 100 notifications/second, ten workers handle 1,000/second.

Queue Partitioning: Split your notification queue into multiple partitions (e.g., by user ID hash or notification type). Each partition has dedicated workers, preventing one hot partition from blocking others. This is like having separate lanes for "express checkout" vs "full cart."

Auto-Scaling: Monitor queue depth metrics. When the backlog exceeds thresholds (e.g., 1,000 unprocessed messages), automatically spin up additional workers. Scale down during quiet periods to save costs.

Load Distribution: Use message routing keys to distribute work evenly. Avoid hot spots where one worker is overwhelmed while others sit idle.

Managing State and Coordination

Keep workers stateless so any worker can process any message. Store shared state (user preferences, delivery status) in distributed caches or databases accessible to all workers.

Key Takeaway: Horizontal scaling of notification workers through parallelization and queue partitioning enables you to handle massive traffic spikes by adding capacity elastically, maintaining consistent latency even as notification volume grows exponentially.