Q30: Cutting Costs on a Database Used Only 48 Hours a Month

Problem Info

Domain	Design Cost-Optimized Architectures
Services	Amazon RDS, RDS Snapshots
Difficulty	Easy–Medium

Question

A team runs resource-intensive tests once a month for 48 hours on a general-purpose RDS for MySQL instance — the only workload on that database. The team wants to reduce testing costs without reducing the instance’s compute/memory attributes.

Which solution is most cost-effective?

Correct Answer

Take a snapshot when testing finishes, then terminate (delete) the DB instance. Restore the snapshot into a new instance the next time testing is needed.

Why this is correct

The core insight: an RDS instance that’s actively used only 48 hours out of roughly 720 hours in a month (under 7% utilization) is being paid for at full price the other 93%+ of the time it’s simply idle.

RDS billing is primarily per running instance-hour. Since RDS instances cannot be “stopped” indefinitely the way EC2 can for long, unattended periods without limitation (RDS auto-restarts a stopped instance after 7 days), and this database sits unused for weeks at a time, the most cost-effective move is to not have a running instance at all between test cycles.

Taking a final snapshot before terminating preserves 100% of the data and configuration — snapshots are billed at a low per-GB storage rate, dramatically cheaper than a running instance. When the next monthly test cycle arrives, restoring the snapshot recreates a new DB instance with the exact same engine, data, and instance class as before — satisfying “without reducing the compute and memory attributes,” since you simply restore to the same instance type used previously.

Why the alternatives fall short

• Use RDS Reserved Instances to lower the hourly rate — Reserved Instances give a discount for committing to run continuously over a 1- or 3-year term; that’s the wrong pricing model for a database that’s only needed 48 hours a month — you’d be pre-paying (or committing) for capacity sitting idle almost all the time.
• Downsize the instance class between tests, then resize back up before testing — resizing does reduce idle-time cost somewhat, but it does change compute/memory attributes during the idle period (and requires manual resize operations before/after every cycle), and still leaves an instance running (and billed) continuously — not as cost-effective as having zero running instance when idle.
• Stop the RDS instance between test cycles — RDS supports stopping an instance to pause compute billing, but AWS automatically restarts a stopped RDS instance after 7 consecutive days, meaning it would auto-restart multiple times before the next monthly test — resuming compute charges unexpectedly and requiring the team to notice and re-stop it repeatedly. A snapshot-and-terminate cycle avoids this entirely.

Exam Tip

“Database (or workload) used only occasionally / a few days a month” + “reduce cost without changing compute/memory specs” → snapshot, then terminate/delete; restore from snapshot when needed again. Remember RDS’s 7-day auto-restart limitation on stopped instances — it’s a frequently tested detail that rules out “just stop it” as a valid answer for month-long idle periods.