Memory hook: Budget warns; quota limits; SLO measures the experience.
Must remember
CapEx purchases long-lived assets; OpEx pays ongoing operating expenses. Cloud often shifts spending toward consumption, but total cost includes staffing, migration, support, data transfer and idle resources. Financial governance assigns ownership, cost allocation, forecasting and review rather than treating the bill as only finance’s problem.
The organization/folder/project hierarchy helps group resources and apply inherited policy. Billing accounts fund linked projects. Budgets warn at thresholds; quotas limit particular resource/API quantities. Neither is a universal automatic hard spending cap. Rightsizing, idle-resource removal, suitable Spot use and commitments for predictable demand address different sources of waste.
High availability uses redundancy to tolerate failures; backups preserve recoverable history; disaster recovery restores service after larger failures. Replication may copy accidental deletion, so it does not replace backup. Recovery needs tested access, dependencies and procedures.
Cloud Monitoring tracks metrics, Logging records events, Trace connects request paths, Profiler identifies runtime hot spots, and Error Reporting groups application errors. Watch latency, traffic, errors and saturation. An SLI is a measurement, an SLO is its target, and an SLA is a contractual commitment. DevOps improves delivery/operation collaboration; SRE applies engineering to reliability and operational toil.
Choose under exam pressure
| Requirement | Choice and reason |
|---|---|
| Warn about projected overspend | Budget notifications with an accountable response process. |
| Recover accidental deletion | Verified backups/version history within retention. |
| Measure user-facing reliability | Meaningful SLIs and SLOs. |
Traps
- Replicas can faithfully replicate mistakes.
- More availability can cost more; choose the target the business needs.
Active recall
1. What is TCO?
Total cost of ownership across direct and indirect lifecycle costs.
2. Is a budget an automatic cap?
No; it provides notification unless a separate control is deliberately implemented.
3. SLI versus SLO?
A measured indicator versus its reliability target.
4. Why test restoration?
A backup is useful only if it can restore required data and service in time.
5. What is saturation?
How close a resource is to a limiting capacity such as CPU, queue or connection capacity.