passdrill

Cost Optimisation

10 cards · AWS SAA-C03 · answer each one, then read the explanation. Your score tallies below.

0 / 10 answered · 0 correct

AWS SAA-C03 · Cost Optimisation · Card 001/010 easy

A team runs a mixed fleet of EC2 instances across several instance families and expects to keep changing instance types as workloads evolve over the next three years. They want the largest possible AWS compute discount without locking themselves into a specific instance family or Region. Which purchase best fits this need?

  1. Purchase EC2 Instance Savings Plans, since they lock in the lowest possible rate for a specific instance family in a specific Region.
  2. Purchase Standard Reserved Instances, since they offer the largest EC2 discount available for a fixed instance type and Region commitment.
  3. Purchase Compute Savings Plans, since the commitment is expressed in dollars-per-hour and automatically applies to EC2 usage regardless of instance family, size, OS, tenancy, or Region (and to Fargate/Lambda usage too).
  4. Stay on On-Demand pricing until the team's instance requirements stabilize, since Reserved Instances and Savings Plans require workloads to be completely static before any discount applies.
AWS SAA-C03 · Cost Optimisation · Card 002/010 medium

A company bought three-year Standard Reserved Instances for m5.large in us-east-1. Midway through the term, they want to switch that commitment to a different instance family, such as c6g.large, without losing the committed value. What is true about this situation?

  1. Because they bought Standard Reserved Instances, they can modify certain parameters (like Availability Zone or instance size within the same family) but cannot exchange them for a different instance family -- only Convertible Reserved Instances can be exchanged for different attributes such as instance family.
  2. They could exchange the Standard Reserved Instances for a different instance family at any time, since all Reserved Instance offering classes support exchanges.
  3. They should have purchased Convertible Reserved Instances instead, but even Convertible Reserved Instances cannot be modified or exchanged once purchased -- only sold on the Reserved Instance Marketplace.
  4. Neither Standard nor Convertible Reserved Instances can be modified or exchanged after purchase; only Savings Plans allow any change to committed compute usage over time.
AWS SAA-C03 · Cost Optimisation · Card 003/010 easy

A fault-tolerant batch-processing application runs entirely on Spot Instances to minimize cost. Its interruption behavior is set to the default (terminate). How much advance warning does EC2 give before reclaiming a Spot Instance, and how can the application detect it?

  1. EC2 gives a 15-minute warning delivered only via email to the account's root user.
  2. EC2 provides no advance warning; Spot Instances can be reclaimed at any moment without notice.
  3. EC2 gives a 2-minute warning, but only if the interruption behavior is set to hibernate; stop and terminate get no warning at all.
  4. EC2 issues a two-minute interruption notice (delivered as an Amazon EventBridge event and as an item in the instance metadata), except when the interruption behavior is hibernate, in which case hibernation begins immediately without that two-minute lead time.
AWS SAA-C03 · Cost Optimisation · Card 004/010 medium

A data lake stores objects with unpredictable, changing access patterns -- some are read constantly, others go untouched for months, and there is no way to know in advance which is which. The team wants automatic storage-cost savings without building lifecycle rules or accepting retrieval fees on frequently-accessed data. Which S3 storage class fits, and how does its cost mechanism work?

  1. S3 Intelligent-Tiering, but objects are moved between tiers only when manually requested via a lifecycle transition rule, and each move incurs a retrieval fee similar to Glacier.
  2. S3 Intelligent-Tiering, which automatically monitors access patterns and moves objects between its frequent-access and infrequent-access tiers with no retrieval fees, charging only a small monthly per-object monitoring and automation fee.
  3. S3 Intelligent-Tiering, which charges the same per-GB rate as S3 Standard at all times, so it is functionally identical to S3 Standard with different marketing.
  4. S3 Intelligent-Tiering, but once an object is moved to a lower-cost access tier, later requests are permanently priced at the higher Glacier retrieval rate regardless of the object's new access pattern.
AWS SAA-C03 · Cost Optimisation · Card 005/010 hard

A company is on the free Basic AWS Support plan and expects Trusted Advisor to flag every idle load balancer, underutilized EBS volume, and Reserved Instance purchase opportunity in the account. What should they understand about Trusted Advisor's cost optimization checks and AWS Support plans?

  1. Every AWS Support plan, including the Basic (free) plan, provides the full set of Trusted Advisor cost optimization checks.
  2. Trusted Advisor's cost optimization checks are informational only and never flag idle or underutilized resources such as unattached EBS volumes or load balancers with no traffic.
  3. The Basic and Developer Support plans give access to only a limited set of core checks; the full set of Trusted Advisor checks -- including cost optimization checks such as idle load balancers, underutilized EBS volumes, and Reserved Instance purchase recommendations -- requires at least a Business-level support plan.
  4. Trusted Advisor's cost optimization checks require a separate paid subscription that is independent of any AWS Support plan.
AWS SAA-C03 · Cost Optimisation · Card 006/010 easy

A solutions architect wants recommendations on right-sizing an EC2 fleet, including identifying idle instances, without manually inspecting each instance's monitoring data. What underlies the rightsizing recommendations that AWS Compute Optimizer produces?

  1. It analyzes each resource's configuration alongside its Amazon CloudWatch utilization metrics (such as CPU and memory usage over recent history) and applies machine learning to generate rightsizing recommendations.
  2. It relies solely on the list price of each instance type, recommending whichever instance is cheapest regardless of actual utilization.
  3. It requires the customer to manually upload utilization spreadsheets, since Compute Optimizer has no direct integration with CloudWatch.
  4. It only produces recommendations for Amazon EC2 instances and cannot analyze Auto Scaling groups, EBS volumes, or Lambda functions.
AWS SAA-C03 · Cost Optimisation · Card 007/010 medium

An application in a private subnet sends a large volume of traffic to Amazon S3 through a NAT Gateway, generating a significant per-GB data processing charge. A cost review recommends adding a Gateway VPC Endpoint for S3 instead. What effect does this have on cost?

  1. A Gateway VPC Endpoint for S3 incurs an hourly charge similar to a NAT Gateway, so switching to a Gateway Endpoint provides no cost benefit.
  2. Only an Interface VPC Endpoint (not a Gateway Endpoint) can be used for Amazon S3, and Interface Endpoints charge the same per-GB processing fee as a NAT Gateway.
  3. Traffic to Amazon S3 from a private subnet must always traverse a NAT Gateway; there is no way to reach S3 without one.
  4. Routing S3 traffic through a Gateway VPC Endpoint avoids sending that traffic through the NAT Gateway at all, and AWS does not charge an additional fee for the Gateway Endpoint itself, eliminating the NAT Gateway's per-GB data processing charge for that traffic.
AWS SAA-C03 · Cost Optimisation · Card 008/010 easy

A team is migrating EBS volumes from gp2 to gp3 primarily to reduce cost while keeping predictable performance. What is a correct statement about how gp3 achieves this compared to gp2?

  1. gp2 volumes let you provision IOPS and throughput independently of the volume's size, while gp3 ties both to how many GiB you provision.
  2. gp3 volumes provide a baseline of 3,000 IOPS and 125 MiB/s of throughput regardless of volume size, and let you provision additional IOPS and throughput independently of size, typically at a lower cost per GB than gp2.
  3. gp2 and gp3 are identical in every respect except that gp3 volumes cannot be used as boot volumes.
  4. gp3 volumes always cost more per GB than gp2 volumes, because gp3 offers a strictly higher maximum IOPS ceiling.
AWS SAA-C03 · Cost Optimisation · Card 009/010 hard

A finance team needs to allocate every dollar of AWS spend down to the individual resource and cost-allocation tag, at hourly granularity, for a chargeback process. Cost Explorer's dashboards aren't detailed enough. Which statement about the AWS Cost and Usage Report versus Cost Explorer is correct?

  1. The Cost and Usage Report delivers the most granular billing data available (down to the individual line-item level, broken down by hour, resource, and tag), updated at least daily to an S3 bucket the account owns, whereas Cost Explorer offers pre-aggregated visualizations and forecasts rather than raw line-item detail.
  2. Cost Explorer and the Cost and Usage Report contain identical data; the only difference is that the Cost and Usage Report has a graphical interface while Cost Explorer only exports CSV files.
  3. The Cost and Usage Report only reports usage at the monthly account-total level, making Cost Explorer the more granular of the two tools.
  4. The Cost and Usage Report cannot be delivered to Amazon S3; it can only be viewed inside the AWS Billing console.
AWS SAA-C03 · Cost Optimisation · Card 010/010 easy

A team is deciding whether raising a Lambda function's configured memory from 512 MB to 1,769 MB will increase their monthly bill, given that the function is CPU-bound and currently runs slowly at low memory. Which statement correctly describes how this choice affects cost?

  1. Lambda charges a flat per-invocation fee regardless of memory or execution duration, so memory configuration has no effect on cost.
  2. Increasing a function's configured memory always increases total cost, because AWS charges strictly by memory size regardless of how long the function runs.
  3. Lambda bills by number of requests plus duration measured in GB-seconds (execution time multiplied by allocated memory), and because CPU power scales with configured memory up to the 1,769 MB point that yields one full vCPU, raising memory can shorten execution time enough to hold flat or even lower the total GB-seconds billed; Compute Savings Plans can also apply to Lambda duration charges the same way they apply to EC2 and Fargate usage.
  4. Lambda usage is billed only in whole-hour increments, similar to EC2 On-Demand Instances, with no per-millisecond granularity.