Compare on-demand, committed/reserved, and spot/preemptible pricing. How do you mix them?
What they are really testing: Whether you understand the discount-for-commitment-or-risk trade-off and can map each model to a workload type (baseline, flexible, fault-tolerant).
A real interview question
Compare on-demand, committed/reserved, and spot/preemptible pricing. How do you mix them?
What most people say
drag me
“Reserved instances are cheaper, so you should use reserved instances.”
Reserved is cheaper only for steady baseline load, committing to capacity you do not consistently use wastes money, and it ignores spot for fault-tolerant work. The answer is a mix matched to workload shape, not one model.
The follow-ups they ask next
What kind of workload is a bad fit for spot instances?
Anything that cannot tolerate interruption: stateful databases, long single-run jobs without checkpointing, latency-critical user-facing services that need guaranteed capacity. Spot suits fault-tolerant/batch work that can be killed and restarted.
What is the risk of over-committing to reserved capacity?
You pay for the commitment whether you use it or not, so committing to more than your stable baseline wastes money if usage drops or you re-architect. Commit only to the reliable, always-on portion.
What the interviewer is listening for
- Maps each model to a workload type
- Knows spot can be reclaimed / for fault-tolerant
- Advocates a mix to usage shape, not one model
What sinks the answer
- "Just use reserved"
- Unaware spot can be reclaimed
- No baseline-vs-burst reasoning
If you genuinely do not know
Say this instead of freezing. Reasoning out loud from what you do know beats silence every single time, and a good interviewer is listening for exactly that.
“On-demand = [full price, no commitment, flexible, for unpredictable/short workloads]. Committed/reserved = [1-3yr commitment for ~30-70% off, for steady baseline]. Spot/preemptible = [up to ~90% off spare capacity that can be reclaimed, for fault-tolerant/batch]. Mix: [commitments for the baseline, on-demand for variable load above it, spot for interruptible/batch], matched to workload shape.”
Keep going with cost
Foundation
What does elasticity mean in the cloud, and how does pay-as-you-go change the cost model compared to running your own servers?
Foundation
What are the main things that drive a cloud bill?
Foundation
Why is data egress (data transfer) often a hidden cost, and how do you reduce it?
Junior
What is right-sizing, and why is over-provisioning the most common source of cloud waste?
Junior
Why is resource tagging important for cost management?
Mid
How do autoscaling and scheduling reduce cost, and what is an easy win for non-production?
Knowing the answer is not the same as recalling it under pressure
Sign in to send the questions you fumble to spaced recall, so they come back right before you would forget them, and learn the concepts behind them with hands-on labs.
Start free