InterviewsVector
AdvancedPrincipalCloud Infrastructure & Platforms60 minutesPro answer

SDV-039

Reserve regional capacity for failover without paying for an idle mirror

Combine reserved floors, reclaimable workloads, provider diversification, and evacuation tests to make standby capacity both economical and credible.

Capacity PlanningDisaster RecoveryCloud CostActive Active

Interview prompt

Problem context

Three regions run near 65% utilization. Losing one requires the other two to absorb traffic, but keeping one-third idle is costly. Batch work and development environments could be evicted, while some databases and quotas cannot scale instantly. Design capacity policy.

Skills being evaluated

failover capacityreclaim policycloud economicsreadiness testing

The full reasoning guide is part of Pro

The scenario and evaluation focus above remain public. Pro unlocks the structured answer, trade-off analysis, follow-up probes, common weak answers, rubric, related reasoning, and any architecture diagram.

Sign in to continue

What the full guide covers

Clarify the decision
Establish scale assumptions
Functional and non-functional requirements
High-level architecture
Data model and flow
Consistency and transaction boundaries
Failure modes and recovery
Security and privacy
Observability and SLOs
Capacity and cost
Alternatives and trade-offs
Evolution and migration
What Staff and Principal candidates should emphasize

Related architecture decisions

Staff · 55 min

Plan regional evacuation with explicit RPO, RTO, and authority transfer

Turn “multi-region” into a rehearsed evacuation protocol with data-loss bounds, capacity, fencing, dependency readiness, and return-home criteria.

Principal · 65 min

Schedule heterogeneous accelerator workloads with fairness and preemption

Balance scarce GPU shapes, topology, queue fairness, checkpointing, reservations, and utilization across interactive and batch workloads.

Senior · 50 min

Reduce infrastructure cost without spending the reliability margin twice

Find waste through unit economics, then protect failure headroom, rollback capacity, and customer SLOs with staged savings and explicit guardrails.

Senior · 50 min

Design an internal platform with paved roads and governed exceptions

Create self-service defaults, golden paths, escape hatches, and product feedback so a platform accelerates teams without becoming a central ticket queue.