AWS Cost and Well-Architected Basics

Two AWS skills separate from all the service-specific ones so far: reading what you actually spent, and choosing a purchasing option before you spend it. Both are measured in this lesson, not asserted.

The Cost Explorer section needs a real AWS account with billing history — there is no LocalStack equivalent for billing data. The pricing comparison is arithmetic on AWS’s published rates and needs nothing but a calculator.

Querying Cost Explorer

aws ce get-cost-and-usage \
  --time-period Start=2026-08-01,End=2026-09-01 \
  --granularity MONTHLY \
  --metrics "UnblendedCost" \
  --group-by Type=DIMENSION,Key=SERVICE \
  --query 'ResultsByTime[0].Groups[].{Service:Keys[0],Cost:Metrics.UnblendedCost.Amount}' \
  --output table

$ aws ce get-cost-and-usage --time-period Start=2026-08-01,End=2026-09-01 --granularity MONTHLY --metrics "UnblendedCost" --group-by Type=DIMENSION,Key=SERVICE --query 'ResultsByTime[0].Groups[].{Service:Keys[0],Cost:Metrics.UnblendedCost.Amount}' --output table
--------------------------------------------------------
|                 GetCostAndUsage                       |
+---------------------------------+----------------------+
|  EC2 - Other                     |  41.2087430000       |
|  Amazon Elastic Compute Cloud    |  70.0812000000       |
|  Amazon Relational Database Svc  |  52.5600000000       |
|  Amazon Simple Storage Service   |  4.1130000000        |
|  AWS Lambda                      |  0.9241000000        |
+---------------------------------+----------------------+

“EC2 - Other” is easy to misread — it’s not idle compute, it’s EBS volumes, snapshots, and data transfer billed under EC2 but broken out separately from instance-hours (“Amazon Elastic Compute Cloud”).

Finding what’s driving the trend

aws ce get-cost-and-usage \
  --time-period Start=2026-06-01,End=2026-09-01 \
  --granularity MONTHLY --metrics "UnblendedCost" \
  --filter '{"Dimensions":{"Key":"SERVICE","Values":["Amazon Relational Database Service"]}}' \
  --query 'ResultsByTime[].{Month:TimePeriod.Start,Cost:Total.UnblendedCost.Amount}'

[
    {"Month": "2026-06-01", "Cost": "26.2800000000"},
    {"Month": "2026-07-01", "Cost": "26.2800000000"},
    {"Month": "2026-08-01", "Cost": "52.5600000000"}
]

Flat for two months, then doubled — this is the shape of “someone enabled Multi-AZ” (the rds lesson’s own example does exactly that), not gradual usage growth. A cost anomaly that doubles cleanly between months is almost always a configuration change, not organic traffic.

The Well-Architected Framework: six pillars

AWS’s free Well-Architected Tool runs a structured questionnaire against a named workload and produces a scored report per pillar — worth running once a workload is real, not for a toy project.

aws wellarchitected create-workload \
  --workload-name "checkout-service" \
  --description "Customer-facing checkout API" \
  --environment PRODUCTION \
  --lenses "wellarchitected" \
  --review-owner "abhishek@acme.example" \
  --aws-regions "us-east-1"

$ aws wellarchitected create-workload ...
{
    "WorkloadId": "a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6",
    "WorkloadArn": "arn:aws:wellarchitected:us-east-1:482719503661:workload/a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6"
}

On-Demand vs Reserved vs Spot: a real comparison

One m5.large in us-east-1, priced three ways, run continuously for a month (730 hours):

† All-Upfront is a single $404.94 payment amortized over the 8,760-hour term, shown here as an effective hourly rate for comparison.

The Reserved discount is a fixed trade of flexibility for price — you commit to m5.large in us-east-1 for a year regardless of whether you keep using it. Spot is priced by real-time supply and demand for AWS’s unused capacity, so $0.0289 is representative, not guaranteed; it can rise, and the instance can be reclaimed with two minutes’ notice.

The decision this table actually drives: a stateless batch worker that can checkpoint and resume belongs on Spot — $21.10 versus $70.08 is a 3.3x cost difference for identical compute. A primary database that must not be interrupted belongs on Reserved or On-Demand, because Spot’s discount is not worth an unplanned two-minute-notice termination on a stateful, hard-to-restart workload.

aws ce get-savings-plans-purchase-recommendation \
  --savings-plans-type COMPUTE_SP --term-in-years ONE_YEAR \
  --payment-option NO_UPFRONT --lookback-period-in-days SIXTY_DAYS \
  --query 'SavingsPlansPurchaseRecommendation.SavingsPlansPurchaseRecommendationDetails[0].{HourlyCommitment:HourlyCommitmentToPurchase,EstimatedSavings:EstimatedMonthlySavingsAmount,SavingsPct:EstimatedSavingsPercentage}'

{
    "HourlyCommitment": "0.142",
    "EstimatedSavings": "31.87",
    "SavingsPct": "31.4"
}

AWS’s own recommendation engine is looking at your last 60 days of actual On-Demand usage — this is a data-driven starting point for a commitment, rather than guessing a number.

Practice

aws ce get-cost-and-usage \
  --time-period Start=2026-08-14,End=2026-09-13 \
  --granularity DAILY --metrics "UnblendedCost" \
  --query 'ResultsByTime[-3:].{Day:TimePeriod.Start,Cost:Total.UnblendedCost.Amount}'

[
    {"Day": "2026-09-10", "Cost": "6.1204"},
    {"Day": "2026-09-11", "Cost": "6.1198"},
    {"Day": "2026-09-12", "Cost": "6.1211"}
]

Reserved has no upfront fee here (No Upfront plan), so it is strictly cheaper per hour
from hour zero: $0.0610 < $0.0960 always. The real break-even question is against
All Upfront: $404.94 upfront ÷ ($0.0960 - $0.0554) saved per hour ≈ 9,975 hours ≈ 13.7
months — longer than the 1-year term, so No Upfront is the better choice unless the
instance will clearly run past a second year too.

10 instances x 4 hours x 22 nights x $0.0289/hr = $25.43/month

aws ce get-cost-and-usage \
  --time-period Start=2026-08-01,End=2026-09-01 --granularity MONTHLY \
  --metrics "UnblendedCost" \
  --filter '{"Dimensions":{"Key":"SERVICE","Values":["Amazon Relational Database Service"]}}' \
  --group-by Type=DIMENSION,Key=LINKED_ACCOUNT \
  --query 'ResultsByTime[0].Groups[].{Account:Keys[0],Cost:Metrics.UnblendedCost.Amount}'

[
    {"Account": "482719503661", "Cost": "52.5600000000"}
]

On-Demand annual:  $70.08 x 12 = $840.96
Reserved annual:   $44.53 x 12 = $534.36
Annual saving:     $840.96 - $534.36 = $306.60  (36.5%)

This closes the AWS track. From here: terraform for infrastructure as code across providers, or data-engineering for how S3 and RDS feed into a broader pipeline.