Resilience Architecture — GCP Professional Cloud Architect (PCA)
Regional and Global Load Balancers Are Not Interchangeable
A scenario requiring 99.99% availability with geographically distributed users draws candidates toward 'any load balancer' as sufficient. The distinction that matters: Global HTTP(S) Load Balancing routes traffic to the nearest healthy backend across regions via Google's anycast frontend — it is the only option that provides cross-region failover within a single configuration. Regional internal load balancers cannot route across region boundaries. Selecting the wrong tier produces a single-region availability ceiling regardless of how many backend replicas exist.
What This Pattern Tests
Resilience questions test whether you match the availability design to the stated SLA. Over-designing wastes money (multi-region for a 99.9% target). Under-designing risks downtime (single-AZ for a 99.99% target). The exam penalizes both.
Decision Axis
Stated SLA determines resilience architecture. Design to the requirement, not to the maximum.
Associated Traps
More Top Traps on This Exam
Decision Rules
Whether the 99.99% availability target and globally-distributed external user base require a regional multi-zone MIG paired with Global HTTP(S) Load Balancing, rather than a simpler and cheaper zonal MIG paired with Regional Internal Load Balancing.
Whether a GKE zonal cluster (single-zone Kubernetes control plane) or a GKE regional cluster (control plane distributed across three zones) is required to satisfy a 99.95% intra-region availability SLA when Cloud SQL HA already provides cross-zone database failover.
Whether the database fault-domain scope (regional single-primary with HA standby versus multi-region synchronous replication) matches the stated five-nines availability target and region-failure tolerance requirement.
Domain Coverage
Difficulty Breakdown