6 / 7 SLA, SLO, SLI & Availability Math

The Average Lied

Your monthly SLI is successful requests divided by total requests, pooled across all regions, and it reports 99.95%. Meanwhile one region, carrying about 2% of traffic, returned errors for six straight hours and its customers are filing tickets. Why did the pooled SLI stay green?

A

Request-ratio indicators report the median customer's experience, so an outage has to affect more than half the user base before the number moves

B

Availability indicators conventionally exclude regional network faults, since those originate outside the service boundary and are not counted against the objective

C

One small region's failed requests are a tiny share of the global denominator, so healthy traffic elsewhere absorbs them and the pooled ratio barely moves

D

Six hours sits below the minimum incident duration a monthly SLO window counts, so a regional event that short never registers against the pooled ratio