AWS RDS Explained — The Multi-AZ Question Everyone Gets Wrong

One RDS question comes up in almost every AWS interview, and a startling number of engineers get it wrong: what does Multi-AZ give you? The wrong answer — the one that comes out automatically — is "it spreads read load across two AZs." It does not. The Multi-AZ standby serves NO traffic. None. It sits there, fully provisioned, costing you exactly double, doing nothing — until the primary fails. Multi-AZ is AVAILABILITY. Read replicas are SCALE. Different features, different problems. They only look the same on a diagram. 9 concepts: 1. Failover — it's a DNS change. Which means your app must reconnect — and the JVM caches DNS FOREVER by default. 2. Read Replicas — replication lag is a CORRECTNESS bug, not a performance one. Read-your-own-writes will bite you. 3. Backups — automated backups die WITH the instance. A restore creates a NEW endpoint. 4. Parameter Groups — static parameters need a reboot, and the console will say "applied" when it isn't. 5. Storage — gp2 burst credits run out SILENTLY. Your DB gets slow, nothing in your code changed. Use gp3. 6. Connections — Postgres spawns a PROCESS per connection. The fix is never to raise max_connections. 7. Security — the "Publicly Accessible" flag, and encryption you cannot enable after creation. 8. Cost — Multi-AZ doubles the bill. Storage only ever grows. 9. The Interview — the five questions, and the answers that land. The line that separates you: the standby serves nothing. Say that in the first 30 seconds and the interviewer knows you've run this, not read about it. #AWS #RDS #Database #PostgreSQL #MySQL #SystemDesign #InterviewPrep #CloudComputing [prepforge:aws-rds]