Database Sharding vs Vertical Scaling for 100GB+ PostgreSQL Tables

When a B2B SaaS reaches millions of rows in a single PostgreSQL table, queries that used to take 50 milliseconds now take 5 seconds. The initial reflex is "Vertical Scaling"—simply paying AWS to upgrade the server's RAM from 16GB to 128GB. Eventually, you run out of hardware upgrades.

The Inevitability of Horizontal Sharding

Instead of building a taller skyscraper, you build multiple houses. Database Sharding slices your massive database logically. For example, all users in Europe are placed securely on a Frankfurt database instance, while US users are placed on an Ohio instance.

When an application query arrives, the networking layer immediately routes it to the specific, lightweight shard. This horizontal distribution guarantees that your latency remains at 10 milliseconds, regardless of whether you have 1 Million or 1 Billion concurrent users.