Regional MIG Design to Survive Zonal Outage at Lowest Cost
Short answer
B Deploy the payment application on a regional MIG so instances spread across all zones in one region, surviving a zonal outage at minimal cost.
Question
A retail company s most critical application is its online payment processing system. The business has a requirement that the system must be able to survive a complete zonal outage while minimizing cost. You need a design solution that can handle a zonal failure. What should you do?
Key Takeaways
The question tests matching the stated failure domain (a single zone) to the smallest design that covers it; the trap is reaching for multi-region active-active architecture that solves a regional outage the business never asked for.
This PCA design question asks for the cheapest architecture that survives a complete zonal outage for a payment-processing application. A regional managed instance group spanning multiple zones in one region satisfies the failure requirement at far lower cost than a multi-region active-active deployment.
Common mistake: Choosing option A (global HTTP(S) load balancer with multi-regional Spanner across two regions), because it feels more 'highly available' — but it over-engineers beyond the zonal-outage requirement and violates the minimize-cost constraint.
Expert Analysis
Why the Answer Is Correct
A regional Managed Instance Group distributes its instances across all zones within one region and, on zone loss, recreates the lost capacity in the remaining healthy zones. That directly satisfies the stated requirement to survive a complete zonal outage, because only the failed zone's VMs are affected while peers in other zones keep serving traffic. Because everything stays inside a single region, you avoid cross-region data egress charges and the premium cost of a multi-regional database, which is what the "minimizing cost" constraint demands. It also keeps the payment system's latency low by keeping compute and data in one region. For a requirement scoped only to zonal failure, this is the minimal-cost design that fully meets the technical requirement.Why the Other Options Are Wrong
Option A over-provisions: an active-active multi-region setup with a global external HTTP(S) load balancer and Spanner protects against an entire region outage, which the business did not ask for, and it carries substantially higher compute, database, and inter-region egress cost — failing the cost objective. Option C substitutes Spot VMs for the base capacity; Spot VMs can be preempted with only short notice, so a payment-processing system cannot rely on them for guaranteed availability, even though a regional MIG would still rebalance. Option D is doubly wrong: multiple regions plus daily snapshots gives only RPO-based recovery measured in hours or days — unacceptable for a critical payment system — and "lowest possible cost" here trades away availability rather than delivering it.Community Comment Notes
Several learners landed on B with sound reasoning; glb2 noted "Answer is B - Regional MIG" and praised it as "Cost-effective solution (not over-engineering)", while Pime13 explained that a regional MIG "automatically spreads VM instances across multiple zones within a single region." Pime13 also correctly flagged that option A "is Over-engineering and Expensive" for a requirement that is only zonal. Gabrycloud leaned toward A, saying "I'm cool with A, unless someone's got a better take based on the docs" — but that comment concedes the point rather than refuting the cost constraint, so the multi-region option remains out of scope for this design.Official Reference
Exam Strategy
When a question names a specific failure domain (zonal, regional, multi-regional), design to exactly that domain and no wider — the cheapest option that covers the stated domain is usually right. Always weigh the explicit cost constraint: extra resilience beyond the requirement is treated as a design failure in PCA questions.
Frequently Asked Questions
Why is the multi-region active-active option A wrong for a zonal-outage requirement?
Option A protects against a full regional outage, which exceeds the stated requirement, and its global load balancer plus multi-regional Spanner costs far more than needed. The minimize-cost constraint rules it out.
Can a regional MIG use Spot VMs to cut cost on this payment system?
No. Spot VMs can be preempted with little notice, so they cannot guarantee availability for a critical payment workload, even though the regional MIG would rebalance capacity.
Discussion 5 contributions
Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded.
From public community threads 5
-
glb2 👍 2 Answered BAnswer is B - Regional MIG. ✔ High availability across zones ✔ Cost-effective solution (not over-engineered)
-
Pime13 👍 1 Answered BZonal Resilience: Unlike a single-zone MIG, a regional MIG automatically spreads VM instances across multiple zones within a single region. If one zone fails, the MIG automatically recreates the lost capacity in the remaining healthy zones. Cost Efficiency: By staying within a single region, you avoid the significant egress costs associated with multi-regional traffic. It also eliminates the need for expensive multi-region database replication (like Spanner) required for active-active cross-region setups. Managed Operations: Google handles the distribution and health checking of instances, minimizing manual administrative toil.
-
Pime13 👍 1A is Over-engineered and Expensive: An active-active multi-region configuration with Cloud Spanner provides massive resilience (regional failure protection), but it is significantly more expensive than a regional setup. Since the requirement is only to survive a zonal outage, this exceeds the requirement and fails the "minimize cost" constraint.
-
0d8df8b 👍 1 Answered BA regional Managed Instance Group runs instances across multiple zones in the same region. So if one zone fails, other zones can still serve traffic.
-
Gabrycloud 👍 1 Answered AI’m cool with A, unless someone’s got a better take based on the docs.
Reader comments & corrections
No reader corrections yet — spotted an error or have a note? Share it below.
Related Analysis
Practice All PCA Questions
Access 81 questions with complete answers and detailed explanations.
View Full PCA Practice Test →