Stage 1 · Lesson 6 of 7
Dependencies and composite availability
A request that needs several components in series can only be as reliable as all of them together: their availabilities multiply. An app at 99.95 per cent, a database at 99.95 and a payment provider at 99.9 give about 99.8 per cent before your own bugs.
Redundancy works the other way: two independent in parallel fail together only when both fail, so two 99 per cent copies give 99.99 per cent. This quick maths tells you whether an is achievable and where redundancy or caching buys the most. It also shows why critical dependencies need their own SLOs.
Common trapassuming replicas fail independently when they share a zone, a configuration or a deployment pipeline, so one bad push takes them all.
A question to check yourself, an AI practice session and 1 cheat sheet entry come with the full roadmap. Unlock
Resources for this stage
Start with· follow one of these
- BookFreeGoogle Site Reliability Engineering Book
Google · Advanced
Free online: read the chapters on embracing risk and service level objectives first
- BookFreeThe Site Reliability Workbook
Google (O'Reilly) · Intermediate
Free online: the Implementing SLOs chapter walks through a worked example and an error budget policy