| Users / Traffic | Database Behavior | Architecture Impact |
|---|---|---|
| 100 users | Single instance DB handles requests easily | Simple architecture, direct DB connection |
| 10,000 users | DB load increases, latency may rise | Introduce connection pooling, caching layers |
| 1 million users | Single DB instance bottleneck; scaling limits reached | Need read replicas, sharding, or NoSQL options |
| 100 million users | Massive data volume and traffic; complex queries slow | Distributed DB clusters, multi-region replication, polyglot persistence |
Why database choice impacts architecture in HLD - Scalability Evidence
The database is usually the first component to break as user count and data grow. This happens because databases have limits on how many queries they can process per second and how much data they can store efficiently. When the DB is overwhelmed, response times increase and the whole system slows down.
- Vertical Scaling: Upgrade to more powerful servers with faster CPUs, more RAM, and SSD storage.
- Read Replicas: Create copies of the database to handle read queries, reducing load on the main DB.
- Sharding: Split data horizontally across multiple database servers to distribute load.
- Caching: Use in-memory caches like Redis or Memcached to serve frequent queries quickly.
- NoSQL Databases: Use databases optimized for specific data types or access patterns (e.g., document stores, key-value stores) for better scalability.
- Polyglot Persistence: Combine multiple database types to handle different parts of data efficiently.
- At 1 million users, assuming 10 requests per user per day, total requests = 10 million/day ≈ 115 requests/sec.
- A single PostgreSQL instance can handle ~5,000 QPS, so DB can handle this load but with little room for growth.
- Storage needs grow with data size; 100 million users may require terabytes of storage.
- Network bandwidth must support data transfer; 1 Gbps network can handle ~125 MB/s, enough for many applications but may need upgrades at large scale.
Start by identifying the database's role and limits in your system. Discuss expected traffic and data growth. Explain how the database choice affects read/write patterns, consistency, and latency. Then describe bottlenecks and propose scaling solutions like replication, sharding, or caching. Finally, mention trade-offs and cost implications.
Your database handles 1000 queries per second (QPS). Traffic grows 10x to 10,000 QPS. What do you do first and why?