SLO & Error Budget Calculator

An availability target is only a commitment once you know what it costs you in minutes, what it gives you as an error budget, and what alerts fire when you start spending it too quickly. Enter the target and the dependency chain behind it — including whether the chain can reach that number in the first place.

Load a scenario

1 · The objective

SLO target (%)

The availability you are committing to, not the one you hope for.

Compliance window

Rolling windows are kinder than calendar months — no budget reset on the 1st.

Requests / sec (average)

Used to express the budget in requests as well as minutes.

Observed availability so far (%)

What you have actually delivered in this window.

Elapsed portion of the window (%)

How far into the window you are — decides whether the current burn rate lands you inside or outside the target.

2 · Can the chain deliver it?

Every dependency on the critical path multiplies. A service is never more available than the product of everything it needs to answer a request. Redundancy assumes independent failure — replicas inside one cluster are not independent, so be honest about the instance count.

	Dependency	Availability (%)	Independent instances

3 · Results

Error budget for this window

Where the unavailability comes from

Share of expected downtime contributed by each dependency, from the numbers above. The widest block is where an availability project should start — not the one that is easiest to fix.

4 · Burn-rate alerts

Alert on how fast the budget is being spent, not on a fixed error rate. Each row needs both its long and short window over the threshold to fire — the short window stops a recovered incident from paging you, the long window stops a brief blip from doing so.

Severity	Budget spent	Long window	Short window	Burn rate	Error-rate threshold	At your traffic

What this means

SLO definition

Composite availability multiplies the published numbers for each dependency, which assumes independent failures. Real systems correlate: replicas of one service in one cluster share a control plane, a node image and a config change, so redundancy inside a single failure domain buys far less than the arithmetic suggests. Treat the composite figure as a ceiling, not a forecast — the achievable number is always lower. Note too that planned work spends the same budget as incidents: node image upgrades, cluster version bumps and certificate rotations all land in it.

Error budget for this window

Where the unavailability comes from

What this means

SLO definitionCopy

SLO definition

SLO Error Budget