Domain 1 of 6 · Chapter 2 of 5

Technical Requirements

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing Professional Cloud Architect premium course — no separate purchase.

Included in this chapter:

  • The six pillars are the lens for technical requirements
  • High availability: redundancy across failure domains
  • Scalability: horizontal and elastic, so capacity tracks load
  • Backup, recovery, and DR: match the pattern to RTO and RPO
  • Gemini Cloud Assist: the architect's collaborator
  • Exam-pattern recognition

DR and HA patterns by RTO/RPO and relative cost

PatternRecovery postureTypical RTOTypical RPORelative cost
Backup and restore (cold)Restore from backups/snapshots into a rebuilt environmentHours (longest)Up to the backup intervalLowest
Warm standbyScaled-down copy running in a second location, scaled up on failoverMinutes to low hoursNear-current (async replication)Medium
Hot / active-activeFull capacity serving in 2+ locations at onceNear zeroNear zero (sync replication)Highest
High availability (in-region)Redundant instances across zones keep one environment servingSurvives component/zone failure, not a region lossNo data loss for in-region failuresAdds redundancy cost only

Decision tree

Continuous serving, orrecover after an outage?High availability:must survive region loss?Disaster recovery:how tight are RTO/RPO?Continuous servingRecover after outageMulti-zone in one regionRegional MIG + LBMulti-regionRedundancy across regionsNoYesBackup and restoreRTO hours, lowest costRPO = backup intervalWarm standbyRTO minutes,async, medium costHot / active-activeRTO/RPO near zero, synchighest costLong RTO / loss OKMinutes RTONear-zero
Cheat sheet

Cheat sheet

  • Map each technical requirement to a Well-Architected pillar first
  • Set realistic reliability targets; tighter ones cost more
  • High availability means redundancy across failure domains, never a bigger machine
  • Multi-zone survives a zone failure; multi-region survives a region failure
  • Use a regional MIG behind a load balancer for multi-zone VM availability
  • Scale horizontally and elastically so capacity follows load
  • A MIG autoscaler needs at least one scaling signal, keyed to what saturates
  • RTO and RPO are different numbers: downtime tolerated vs data loss tolerated
  • Pick the cheapest DR pattern that still meets RTO and RPO
  • Backup-and-restore can never deliver a near-zero RPO
  • Synchronous replication buys a near-zero RPO; async trades a small window for distance and cost
  • High availability and disaster recovery are distinct requirements
  • An untested recovery plan is not a recovery plan
  • Gemini Cloud Assist is a tool for the architect, not a tier in the architecture
  • Place compute and data near users to meet a latency budget
  • Autoscaling serves reliability and cost optimization at the same time
  • Regional Persistent Disk replicates synchronously across two zones for near-zero RPO
  • Tune the MIG autoscaler with initialization period, scale-in controls, and predictive scaling
  • Use aggressive health checks for load balancing, conservative ones for autohealing
  • Roll out a new model by splitting traffic on one Vertex AI endpoint

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. Google Cloud Well-Architected Framework Well-Architected
  2. Well-Architected Framework: Reliability pillar Well-Architected
  3. Well-Architected Framework: Performance optimization pillar Well-Architected
  4. Compute Engine: Regions and zones
  5. External Application Load Balancer (HTTPS load balancing)
  6. Autoscaling groups of instances
  7. Architecting disaster recovery for cloud infrastructure outages: Planning guide Well-Architected
  8. Disaster recovery planning guide (building blocks) Well-Architected
  9. Gemini Cloud Assist overview