GCP Professional Cloud Architect Practice Question 9: SLI SLO Error Budget Definition

This is the standard shape of a well-formed reliability target. The SLI is a request-based availability indicator, successful responses over all responses, measured at the serving edge, which is the closest practical proxy for what callers experience. The SLO states the desired value of that SLI over a compliance period; a rolling window continuously evaluates the last *n* days so that a bad week keeps mattering until it ages out, rather than being forgiven by a calendar boundary. The error budget follows arithmetically as (1 − SLO) × eligible events, and turning it into a written policy is what gives the organization the agreed stop-shipping rule it asked for.

Option A measures a cause, not the customer experience. CPU utilization is a system-level signal that can look perfectly healthy while callers receive errors from a downstream dependency, and it generates no error budget because there is no notion of a failed request.

Option C sets a target that cannot be met and cannot be reasoned about. A 100% SLO leaves an error budget of zero, so every deploy is a violation and the concept loses its power to arbitrate priorities; paging on each individual 5xx also guarantees alert fatigue rather than reliability.

Option D measures the probe, not the service. A single synthetic check from one region samples a narrow path at low frequency, so it misses failures affecting real traffic and reports false outages when the probe's own region has trouble, far too coarse to defend a 99.99% claim.

Reference: SLI, SLO, and error budget concepts