SD10.1 · LESSON 1 OF 6

Indicators and objectives

Define service objectives and use telemetry, failure handling and recovery procedures to protect user outcomes.

THE IDEA

Indicators and objectives

A service level indicator (SLI) measures an outcome users care about, such as the fraction of requests that succeed. A service level objective (SLO) sets a target for it over a window. Choose the denominator on purpose: a time-based 99.9% over 30 days (43,200 minutes) allows 43.2 bad minutes, while a request-based 99.9% over 10 million requests allows 10,000 failed requests. The two are not interchangeable.

nines (3 = 99.9%)99.9%window (days)30requests in the window10,000,000

window = 30 × 1440 = 43,200 minutes time-based SLO 99.9%: allowed bad time = 43.2 minutes request-based SLO 99.9% of 10,000,000 requests: 10,000 failed requests allowed the two use different denominators: do not read one as the other

An SLI measures a defined outcome; an SLO is a target for it over a window. Say whether you count requests or time, what counts as good, and over how long.

Change the target, window and traffic.

Predict first. Can a monthly request success ratio be silently interpreted as minutes of outage?

Chapter context

Reliability is measured against what users need, not whether every component reports healthy. Observability connects symptoms to evidence, while resilience and recovery determine what happens when dependencies or infrastructure fail.

TRY IT YOURSELF

Part 1. Do it hands-on

Work out the allowances by hand, then write the indicator helpers.

nines (3 = 99.9%)99.9%window (days)30requests in the window10,000,000

An SLI measures a defined outcome; an SLO is a target for it over a window. Say whether you count requests or time, what counts as good, and over how long.

bad minutes at 99.9% (30 d) at 99.99% failed requests allowed, 10M at 99.9% request ratio = minutes of outage? (yes/no)

To do: A 30-day window has 43,200 minutes. A time-based availability SLO of 99.9% allows how many bad minutes (1 decimal)?
To do: At 99.99% (2 decimals)?
To do: A request-based SLO of 99.9% over 10,000,000 requests in the window allows how many failed requests?
To do: Can a monthly request success ratio be silently interpreted as minutes of outage? Type yes or no

Part 2. Do it in code

allowed_bad_minutes(slo, days): (1 − slo) × days × 1440.
allowed_bad_requests(slo, total_requests): (1 − slo) × total_requests, rounded to the nearest integer.
request_sli(good, total): good / total (1.0 when total is 0 — nothing failed).
time_sli(bad_minutes, days): 1 − bad_minutes / (days × 1440).
meets_slo(sli, slo): sli >= slo (with a 1e-12 tolerance for rounding).
same_unit(kind_a, kind_b): the two indicators are comparable only if they are the same kind ("request" or "time"); raise ValueError for anything else, return a == b otherwise.

def allowed_bad_minutes(slo, days):
    pass

def allowed_bad_requests(slo, total_requests):
    pass

def request_sli(good, total):
    pass

def time_sli(bad_minutes, days):
    pass

def meets_slo(sli, slo):
    pass

def same_unit(kind_a, kind_b):
    pass

Ctrl/⌘ + Enter runs it

Stuck, or something not working? You can come back to it any time.

CHECK YOUR UNDERSTANDING

Can a monthly request success ratio be silently interpreted as minutes of outage?

ANoBYes

Chapter reference: math, worked example & code

This section belongs to the whole chapter, so it is the same on every lesson in it and may cover a different lesson's topic. The code for the experiment you just ran is inside the experiment itself.

Math and notation

For a time-based target A over W minutes, allowed bad time is $(1-A)W$ . A simplified two-replica failover availability model is $1-(1-a)^2$ only if failures are independent and failover is perfect. It excludes many shared-dependency and detection failures.

Worked example

A 30-day time window contains 43,200 minutes. At 99.9% time-based availability, allowed bad time is 43.2 minutes; at 99.99%, it is 4.32 minutes. Two independent 99%-available replicas with ideal failover give 99.99% in the toy model, not a general deployment guarantee.

Reference code

The following example uses the Python standard library and does not call a remote service.

from math import isclose
minutes=30*24*60
assert minutes == 43200
assert isclose((1-.999)*minutes,43.2)
assert isclose((1-.9999)*minutes,4.32)
assert isclose(1-(1-.99)**2,.9999)

Python examples are provided for study and local execution; this page does not execute Python.

Common mistakes

Availability and durability differ. An average is not a latency percentile. Backup existence is not proof of recoverability. Correlated failures can dominate replica calculations.