Cloud Cost Optimization Flashcards

Question 1

What is the pay-as-you-go principle in cloud computing?

Accepted Answer

Pay-as-you-go means customers are billed only for the resources they actually use, usually measured by time, capacity, or requests, rather than paying upfront for fixed hardware. Charges usually stop when resources are stopped or deleted, which is why unused or forgotten resources can still accumulate costs. This is where most optimization work begins.

Question 2

What are the main components that make up a cloud bill?

Accepted Answer

A cloud bill is usually built from compute, storage, network data transfer, and managed services, each billed by its own unit such as instance hours, gigabytes stored, or gigabytes transferred out. Many services also add charges for requests, licenses, or support plans. Sorting the bill by service shows where spending concentrates.

Question 3

Why is data egress often a surprise item on cloud invoices?

Accepted Answer

Data egress is the cost of moving data out of a cloud provider's network, such as to the internet or to another cloud. Providers usually charge per gigabyte for this traffic, while inbound transfer is generally free. Architectures that move large volumes between regions or services can therefore produce surprising charges.

Question 4

What does a unit price mean in cloud pricing?

Accepted Answer

A unit price is the charge for one measurable quantity of a service, such as one hour of a virtual machine, one gigabyte of storage per month, or one million API requests. Total cost equals unit price multiplied by the quantity consumed. This lets engineers estimate how a design change affects the bill.

Question 5

What is a billing account or subscription used for?

Accepted Answer

A billing account is the top-level container that receives invoices and pays for all resources beneath it. Subscriptions or projects often sit underneath, letting organizations separate spending by team, product, or environment. Clear structure at this level makes it much easier to see which group is responsible for each charge.

Question 6

What is a cost allocation report and why is it useful?

Accepted Answer

A cost allocation report breaks total spending into groups such as departments, applications, or customers, based on tags or account structure. It shows who is responsible for each part of the bill. Finance and engineering teams use it to hold owners accountable, compare efficiency between services, and decide where optimization effort pays off.

Question 7

Why do cloud costs tend to rise over time without active management?

Accepted Answer

Costs tend to grow because usage expands as products grow, while forgotten resources, oversized instances, and duplicate environments linger after projects end. Prices can also change, and new services are adopted without review. Without regular attention, small inefficiencies compound into large monthly charges that nobody explicitly approved.

Question 8

What is cost visibility, and why must it come before optimization?

Accepted Answer

Cost visibility means having accurate, timely information about who spends what, on which services, and for what purpose. Teams cannot improve spending they cannot see. Accurate data identifies the largest drivers, reveals anomalies, and shows whether changes actually worked, so optimization decisions rest on evidence rather than guesses.

Question 9

What is on-demand pricing, and when is it the right choice?

Accepted Answer

On-demand pricing charges the standard rate for resources with no upfront commitment and no minimum term. It is the right choice for unpredictable, short-lived, or newly launched workloads, where flexibility matters more than discounts. Because it carries the highest unit rates, steady workloads usually cost less under commitment plans.

Question 10

How do reserved instances or savings plans reduce cloud costs?

Accepted Answer

Reserved capacity and savings plans trade a commitment to a steady level of usage over a fixed term for lower rates than on-demand pricing. The discount applies when usage matches the commitment. If usage drops, you still pay for the committed amount, so commitments should cover only the steady baseline.

Question 11

What is a spot instance, and what is its main trade-off?

Accepted Answer

A spot instance is spare cloud capacity offered at a steep discount, but the provider can reclaim it with little notice when demand rises. The main trade-off is lower cost in exchange for possible interruption. Spot capacity suits fault-tolerant work such as batch processing, rendering, or stateless workers that can checkpoint and restart.

Question 12

Why should commitments be based on a baseline rather than peak usage?

Accepted Answer

A baseline is the minimum level of usage that persists around the clock, so committing to it is nearly always used. Committing to peak usage leaves capacity idle during quieter periods, which wastes the discount's value. Covering peaks with on-demand or spot capacity keeps the commitment fully utilized.

Question 13

What does utilization mean for a committed discount?

Accepted Answer

Utilization is the percentage of a commitment that actually gets consumed during its term. High utilization means the discount is delivering savings, while low utilization means money is spent on capacity that sits unused. Teams should monitor utilization regularly and adjust future purchases when patterns shift.

Question 14

How do commitment-based discounts differ from pay-as-you-go pricing?

Accepted Answer

Commitment-based discounts require a promise to spend or use a minimum amount over a set term, in return for lower rates. Pay-as-you-go pricing has no promise, so bills rise and fall with usage, but unit rates stay higher. The choice depends on how predictable your demand is and how much flexibility you need.

Question 15

What is a reservation coverage rate and why track it?

Accepted Answer

Coverage rate is the share of eligible usage that is paid for at discounted commitment rates instead of on-demand prices. Tracking it shows whether purchases are keeping pace with growth. A falling coverage rate may signal that new workloads are running at full price, which is a common source of avoidable spend.

Question 16

Why should you compare pricing options before selecting a region or instance family?

Accepted Answer

Prices for identical resources can differ across regions and instance families, sometimes substantially. Comparing options early lets teams choose the lowest-cost configuration that still meets latency, compliance, and performance needs. Changing these choices later often requires migration work, so the decision is cheaper to get right at the start.

Question 17

What does right-sizing mean in cloud cost optimization?

Accepted Answer

Right-sizing means matching the capacity of a resource, such as its CPU, memory, or storage, to what the workload actually needs. Many resources are provisioned generously for safety and then run at low utilization. Adjusting them to observed demand, rather than initial guesses, typically reduces cost without hurting performance.

Question 18

Which utilization metrics best help identify oversized compute instances?

Accepted Answer

CPU utilization, memory utilization, network throughput, and disk input and output are common signals. Look at sustained averages and peak values over a representative period, such as several weeks, rather than a single snapshot. An instance that stays under its limits across normal cycles is a strong candidate for a smaller size.

Question 19

Why is memory utilization sometimes overlooked during right-sizing?

Accepted Answer

Many teams focus on CPU because it is easy to graph, but memory pressure can cause slowdowns, swapping, or crashes that CPU charts miss. Some platforms do not report memory use by default, so it requires extra agents or metrics. Ignoring it can lead to downsizing that hurts reliability, undoing the savings.

Question 20

What is the risk of cutting resources too aggressively?

Accepted Answer

Cutting too far can cause performance degradation during traffic spikes, failed batch jobs, or slower recovery after failures. Savings from a single resource can be erased by incidents, support time, and customer impact. A safer approach is to reduce size gradually, monitor closely, and keep a quick path to scale back up.

Question 21

How do autoscaling policies help improve resource efficiency over time?

Accepted Answer

Autoscaling adds capacity when demand rises and removes it when demand falls, so the fleet tracks load rather than staying at peak size. Well-configured policies use metrics such as CPU, queue depth, or request rate to trigger changes. The result is capacity that follows demand rather than peak size all day.

Question 22

Why should you schedule nonproduction environments to shut down?

Accepted Answer

Development, testing, and staging environments rarely need to run overnight or on weekends, yet they often run continuously by default. Scheduling them to stop outside working hours can cut their compute costs substantially with little effort. Teams should also confirm that data and configuration persist safely when the environments are paused.

Question 23

What is an idle resource, and how do teams find them?

Accepted Answer

An idle resource is something still billed but doing little or no useful work, such as an unattached disk, an unused public address, or a server with near-zero traffic. Teams find these using provider recommendations, tag reports, and utilization reviews. Confirm each candidate with its owner before deleting it.

Question 24

Why do storage tiers and lifecycle policies save money?

Accepted Answer

Cloud storage offers tiers with different prices, where colder tiers cost less per gigabyte but charge more to retrieve data. Lifecycle policies move objects to cheaper tiers or delete them as they age. This matches storage cost to how often data is accessed, which is usually infrequent for old logs and backups.

Question 25

What is the benefit of managed services for cost optimization?

Accepted Answer

Managed services move operational work, such as patching, scaling, and backups, to the provider, which can reduce staff time and idle capacity. However, managed offerings may carry higher unit prices than self-managed equivalents. The right comparison includes labor, reliability, and maintenance effort, not just the instance rate.

Question 26

Why do serverless platforms often reduce costs for spiky workloads?

Accepted Answer

Serverless platforms bill for execution time and requests rather than for servers that sit idle between events. For workloads with sharp, infrequent bursts, this can be cheaper than keeping capacity running at all times. The savings shrink, however, for steady high-volume workloads, where an always-on server may be cheaper per unit of work.

Question 27

What is data locality, and how does it affect transfer costs?

Accepted Answer

Data locality means placing compute close to the data it processes, ideally in the same region or zone. When services in different regions or availability zones exchange large volumes, transfer charges can add up quickly. Co-locating tightly coupled components reduces that traffic and often improves latency at the same time.

Question 28

How can caching reduce cloud spending on compute and data transfer?

Accepted Answer

Caching stores frequently requested results closer to users or in memory, so fewer requests reach expensive databases, functions, or origin servers. Fewer requests mean less compute time and less data leaving the network. Cache hit rates determine how much value a cache delivers, so teams should measure them before expanding capacity.

Question 29

What is a content delivery network and how does it save money?

Accepted Answer

A content delivery network is a distributed set of servers that caches static and streamed content at edge locations near users. Serving content from the edge reduces traffic from origin servers, which can lower both compute load and outbound data charges. Edge delivery is often cheaper per gigabyte than serving from origin storage.

Question 30

Why do containers and higher density packing reduce infrastructure costs?

Accepted Answer

Containers package applications with only their dependencies, so several can share one host with less overhead than separate virtual machines. Higher packing density means fewer nodes must be paid for to run the same workload. Scheduling tools that place workloads efficiently, combined with accurate resource requests, make this consolidation practical and safe.

Question 31

How does infrastructure as code support cost control?

Accepted Answer

Infrastructure as code defines resources in version-controlled templates, so environments can be created consistently and removed reliably. This makes it easier to enforce approved instance sizes, required tags, and standard shutdown schedules. Because changes are reviewed and recorded, cost-relevant decisions become visible before they reach the bill.

Question 32

Why should teams consider cost during architecture reviews?

Accepted Answer

Design decisions lock in cost patterns that are expensive to change later, such as data flows, storage formats, and service choices. Including cost questions in reviews surfaces trade-offs while options are still open. Asking how a design scales and what it costs at peak helps teams avoid expensive surprises before they are built.

Question 33

What is resource tagging, and why is it essential for cost work?

Accepted Answer

Resource tagging attaches key-value labels, such as owner, team, application, or environment, to cloud resources. Tags let cost reports group spending in meaningful ways, showing which teams or products drive each charge. Without consistent tags, much spending becomes unassigned, and nobody feels responsible for reducing it.

Question 34

How should a team set a cloud budget and alerts?

Accepted Answer

Set a budget for each account, project, or team based on forecasted demand, then configure alerts at several thresholds, such as fifty, eighty, and one hundred percent of the monthly amount. Early warnings give owners time to act before overruns occur. Budgets work best when reviewed monthly and adjusted as products change.

Question 35

What is an anomaly alert in cloud cost monitoring?

Accepted Answer

An anomaly alert notifies teams when spending deviates sharply from its normal pattern, even if the total remains under budget. Sudden spikes often reveal misconfigurations, runaway jobs, leaked credentials, or unexpected traffic. Catching them within hours rather than at the end of the month can prevent large losses.

Question 36

Why do showback and chargeback models change team behavior?

Accepted Answer

Showback reports show teams what they spend, while chargeback actually transfers those costs to their internal budgets. Making spending visible to owners encourages them to question waste and weigh trade-offs they previously ignored. Both approaches work best when cost data is accurate, understood, and paired with clear guidance on what teams can change.

Question 37

What does a cost optimization review cadence involve?

Accepted Answer

A review cadence is a regular schedule, often monthly, where owners examine spending trends, top cost drivers, anomalies, and progress on planned savings. Each review should end with specific actions, owners, and deadlines. Consistent rhythm turns optimization from a one-time cleanup into an ongoing habit.

Question 38

How do guardrails and policies prevent expensive mistakes?

Accepted Answer

Guardrails are preventive rules that restrict choices, such as blocking very large instance types, limiting regions, or requiring approved configurations. Policies can also require tags before a resource is created. Preventing a costly choice is usually cheaper than detecting and correcting it afterward, and it reduces the need for cleanup work.

Question 39

What makes a cost metric useful for decisions?

Accepted Answer

A useful cost metric is tied to business output, such as cost per transaction, cost per active customer, or cost per build. Raw totals rise with growth even when efficiency improves, so ratios reveal whether spending is becoming more or less productive. Good metrics are simple, consistent, and reviewed alongside the goal they measure.

Question 40

How can teams measure whether an optimization actually saved money?

Accepted Answer

Compare cost before and after the change, adjusted for changes in traffic, volume, and pricing, so growth is not mistaken for savings. Track the same unit metrics over a comparable period. Record the expected savings beforehand so results can be checked against the original forecast.

Cloud Cost Optimization Flashcards

1Cloud billing fundamentals

2Pricing models and commitments

3Right-sizing and resource efficiency

4Architecture and automation savings

5Governance, tagging, and monitoring

Learning Resources