RousselTMLEARNING

Observability

SRE: SLI, SLO and SLA explained simply

The fundamentals of Site Reliability Engineering (SRE): the difference between SLI, SLO and SLA, choosing indicators, number of nines and allowed downtime.

1. Why SRE? #

Measuring reliability from the user's point of view.

The core idea #

SRE (Site Reliability Engineering), formalized by Google, starts from an observation: 100% reliability is neither possible nor desirable. Each extra nine is very expensive and slows down delivery, while users cannot tell the difference between 99.99% and 100% (their own network is less reliable).

So you set an explicit reliability target, measured from the user's point of view, and accept the remaining margin as a budget to spend (deployments, experiments, incidents).

2. SLI, SLO, SLA #

Three acronyms not to confuse.

Definitions #

SLI, SLO and SLA
The SLO is stricter than the SLA, so you can react before paying penalties.
TermDefinitionExample
SLI (indicator)A measure of service quality, usually a ratio of good events to total eventsShare of HTTP requests without a 5xx error and served in less than 300 ms
SLO (objective)The target value of the SLI over a time window99.9% over a rolling 30 days
SLA (agreement)A contractual commitment with consequences99.5% per month, otherwise a 10% discount

The most common SLIs #

Type of serviceRecommended SLI
API, websiteAvailability (non-5xx responses) and latency (share of requests under a threshold)
Batch processingFreshness (age of the last successful run), coverage
Data pipelineFreshness, correctness
StorageDurability, throughput
Good practice: Express latency as a ratio (« 99% of requests in less than 300 ms ») rather than an average or an instantaneous percentile: the ratio accumulates over the window and is computed exactly with histogram buckets.

3. How many nines? #

The objective sets the budget.

Allowed downtime #

Allowed downtime
Going from 99.9% to 99.99% divides the tolerated incident time by ten.
SLOPer 30 daysPer year
99%7 h 12 min3 d 15 h
99.5%3 h 36 min1 d 19 h
99.9%43 min8 h 46 min
99.95%21 min 36 s4 h 23 min
99.99%4 min 19 s52 min
Common mistake: An SLO stricter than the reliability of its dependencies is unsustainable: a service that depends on a 99.9% database cannot aim for 99.99% without redundancy.

Related articles

Time-based vs event-based SLO: understanding, computing and choosing (with the Qonto example)

Differences between time-based SLOs (time slices, uptime) and event-based SLOs (good events over total): defi…

Chaos engineering: principles, tools and practice with SLOs

A complete guide to chaos engineering: principles, steady state and hypothesis, blast radius, link with SLOs …

The error budget: making decisions with SLOs

Calculate and track an error budget, write an error budget policy and use it to arbitrate between new feature…

Keep learning

Support us

♥ SUPPORT US

Support RousselTM Learning

Courses, articles, labs and certificates stay free with no intrusive ads. A voluntary contribution, of any amount you choose, pays for hosting, lab environments and new content.

Contribute with PayPal Any amount · card or PayPal accountLearn more
QR code PayPal
Or scan this QR code with your phone

Thank you, every contribution counts!