Emergency Capacity Reservation Decision

A major demand event is approaching, and the window to lock in reserved (committed, non-refundable) capacity at guaranteed pricing is closing. The demand forecast is genuinely uncertain — it comes as a band (p50/p90/p99), not a single number — and the business depends on deciding how much capacity to reserve, and where, before the deadline passes. Five business invariants govern every decision: (1) Critical-Tier workload must have enough protected capacity under the authored demand band; (2) Optional-Tier workload may use on-demand capacity or be degraded — it is never required to be fully protected; (3) every reservation has real, non-refundable cost whether or not the event's actual demand ever uses it; (4) reserved capacity must stay distributed across regions inside this business's resilience policy — concentrating it in one region creates a hidden risk even when the aggregate total looks adequate; (5) the forecast's own stated uncertainty must be treated explicitly in the plan, never collapsed to a single expected-case number. Decide how much capacity to reserve in each region, whether to enable a bounded on-demand fallback for anything reserved capacity doesn't cover, whether to deliberately shed Optional-Tier workload under pressure to protect Critical-Tier capacity, which percentile of the forecast band the plan is actually sized against, and — critically — whether locking in the maximum possible reservation everywhere is genuinely the safest choice, or a trap that pays for protection the forecast's own worst case will never need.

Difficulty

Principal

Expected Time

~25 min

Category

Infrastructure & Capacity