Cloud Economics and SLA Questions and Answers

Answer: A

Correct answer: it separates growth in usage from genuine inefficiency

A total bill that rises tells you nothing on its own, since a business serving twice as many customers should cost more. Dividing by transactions, users or orders gives a unit figure that stays flat when growth is the cause and rises when something has become wasteful. It is the one metric that lets an engineering team be held to account for cost without being penalised for success.

Answer: B

Both are true and the reason gives the mechanism, and this option value is often worth more than the headline price comparison. An experiment that fails cheaply is what makes running many experiments sensible in the first place.

Answer: C

Correct answer: A-2, B-4, C-1, D-3

The four models trade flexibility against price. On demand costs most per unit and commits to nothing. Committed use exchanges a one or three year term for a substantial discount. Spot capacity sells the provider’s spare fleet cheaply on the understanding that it may be taken back. And the free tier is a small allowance intended for evaluation and learning rather than production.

Answer: D

Correct answer: the provider may reclaim it at short notice

Spot instances are the provider’s unused capacity, sold at a steep discount because it can be withdrawn - typically with a warning of a couple of minutes - whenever a full-price customer wants it. The hardware is the same as any other instance; what differs is the guarantee, and the discount is payment for accepting that interruption.

Answer: A

Correct answer: the primary database of a payments system

A primary transactional database cannot tolerate being reclaimed with two minutes’ notice, and its state cannot simply be recreated. The other three all share the property spot capacity requires: work that can be interrupted and resumed or retried without loss, whether through checkpointing, a retry queue or replication across a fault-tolerant cluster.

Answer: B

Both are true and the reason gives the risk, and a commitment sized to a workload that later shrinks costs more than on demand would have. Sizing the commitment to the reliable baseline rather than to the peak is the usual advice.

Answer: C

The first two are correct and together give the sizing rule. The third ignores that unused commitment is pure waste, which is why the baseline rather than the peak is the right thing to commit. Commitments are also usually layered, with more added as the baseline proves itself.

Answer: D

Correct answer: data leaving the provider's network towards the internet

Egress is outbound traffic to the internet, and it is charged; inbound traffic is generally free, and data at rest and traffic between disks on one instance are not transfer at all. The asymmetry is deliberate and consequential: it is cheap to put data into a cloud and expensive to take it out, which is a real component of the lock-in discussed elsewhere in this subject.

Answer: A

Both are true and the reason gives the principle, and it explains why analytics is usually run beside the storage it reads. The same reasoning discourages designs that shuttle large datasets between providers routinely.

Answer: B

Correct answer: False

Cross-zone traffic within a region is metered on the major providers, in both directions. The charge is small per gigabyte but the volumes can be large, and a chatty application spread across zones for resilience can accumulate a surprising bill. It is why the usual pattern is to keep the chatty components together within each zone and run the whole stack in several zones.