Why is database connection pooling important, especially at scale?

Question

Accepted Answer

Database connections are not cheap. Each one costs memory and setup time, and in Postgres every connection is backed by its own server side process using several megabytes, and there is a hard cap, max_connections, which defaults to 100. If the application opens a connection per request, then under real load with a few thousand concurrent requests you blow past that limit in seconds, and the database starts refusing connections or grinds as it thrashes between too many backends, which can take the whole system down. A connection pool fixes this by keeping a small fixed set of warm, established connections, say 20 per instance, and lending them out and taking them back, so thousands of requests share a bounded number of connections and nobody pays the open and close cost each time. The problem gets acute in serverless or heavily horizontally scaled environments, because every function instance or pod wants its own pool. 50 pods times a pool of 20 is 1,000 connections, which is 10 times what the default Postgres will take. The fix there is an external pooler or proxy, PgBouncer in transaction mode or something like RDS Proxy, sitting between the app and the database and multiplexing many client connections onto a bounded number of real ones. Pooling is really about treating connections as a scarce resource and capping how many ever reach the database.

Answer

Why is connection management especially hard in serverless?

Each function instance scales independently and may hold its own connections, so a spike spawns far more connections than the DB allows. You need an external pooler/proxy (PgBouncer, RDS Proxy) to bound total connections.

Answer

What happens when you hit max_connections?

New connection attempts are refused (errors) and existing work slows as the DB struggles. Pooling bounds connection count so you stay under the limit even under load.

Why is database connection pooling important, especially at scale?

Why is database connection pooling important, especially at scale?

The follow-ups they ask next

What the interviewer is listening for

What sinks the answer

If you genuinely do not know

Keep going with databases

Knowing the answer is not the same as recalling it under pressure