INFRASTRUCTURE LEARNINGINFRASTRUCTURE LEARNING
インフラ大全インフラ大全IT INFRASTRUCTURE / LEARNING
Infrastructure learning guide
Find a lesson or lab/

Esc to close · Search terms stay on your device

LEARNING日本語

Topic lessonTopic: Monitoring and recovering event processing

Calculate an SLO and error budget from completed work

Define a good outcome before setting a service-level objective. Use an image-processing example to calculate how many failures the target allows and how fast that allowance is being consumed.

What you will learn

Choose the denominator, success condition, and observation window for a request-based SLI, then calculate the remaining error budget.

Prerequisite

All observations are fictional. This example covers operational objectives, not service agreements or refunds.

Measure the completion the user needs

A request-based SLO measures the fraction of eligible requests that qualify as good. Here an image conversion is good if its result is stored within the agreed deadline after intake.

Turn a user outcome into a measurable objective
Turn a user outcome into a measurable objective1: User outcome — Choose the operation — Population. 2: Population — Define numerator and denominator — Success condition. 3: Success condition — Compare with the target — Comparison123User outcomePopulationSuccess conditionComparisonTurn a user outcome into a measurable objective1: User outcome — Choose the operation — Population. 2: Population — Define numerator and denominator — Success condition. 3: Success condition — Compare with the target — Comparison123User outcomePopulationSuccessconditionComparison
  1. User outcome → PopulationChoose the operation
  2. Population → Success conditionDefine numerator and denominator
  3. Success condition → ComparisonCompare with the target
User outcome
Use the converted image on time
Population
Requests accepted during the window
Success condition
Result stored within two minutes
Comparison
Attainment and remaining allowance

Calculate a 99.5% target for 10,000 requests

Fictional requests whose two-minute outcome windows are complete
MeasureCountInterpretation
Eligible requests10,000Denominator includes the observed eligible work.
Allowed failures at 99.5%5010,000 × 0.005.
Failures or missed deadlines3099.7% attainment, leaving the equivalent of 20 requests.

The remaining 20 is a calculation for the observed volume. A moving window changes its contents when requests arrive and expire. Compare a short-window failure rate with the allowed rate to track how quickly the budget is being consumed.

Account for blind spots and short-term changes

Separate intake responses from completed work. An acceptance response does not establish that an output was stored. Correlate results with deadlines and decide how to handle gaps in measurement.

Keep burn rate and budget remaining distinct. A short-window failure rate of 1.5% divided by an allowed rate of 0.5% gives a burn rate of 3. Use the longer observation window separately to assess attainment.

Use matching conditions for targets and observations

Choose an option to see the result and explanation.

What counts in the denominator when measuring successful storage?
Read the answer and reason

Answer: Accepted requests that meet the population definition · Count the eligible requests, including those that failed to store a result.

There are 10,000 requests, a 99.5% target, and 30 failures. How much budget remains?
Read the answer and reason

Answer: The equivalent of 20 requests · Subtract 30 observed failures from 50 allowed.

The short-window failure rate is 1.5% and the allowed rate is 0.5%. What follows?
Read the answer and reason

Answer: The budget is burning at three times the allowed rate · The ratio is 3 when measured on the same basis.

Check, then explain in your own words

  1. Define eligible requests, the deadline, and the good outcome.
  2. Calculate attainment and remaining budget from the denominator and failures.
  3. Compare short-window burn with long-window budget.
Check your understanding: Can the SLO be missed even when every intake succeeds?

Yes. If it measures on-time storage, too many failed or late outputs can miss the target.

An intake response and the outcome the user needs are separate measurements.

Sources and conditions checked

The official references used in this lesson are collected here with their scope, conditions and corresponding sections. Independent expert review and live AWS comparison have not been performed.