Cross-Region Replication Cost Explosion
A replication-policy change applied last week widened cross-region fan-out and switched every data class — critical, standard, and low-priority alike — to synchronous replication across US, EU, and APAC. Freshness is excellent across every class and nothing is currently failing, but cross-region transfer cost, replication fan-out, and storage-duplication cost have all risen sharply and show no sign of coming back down on their own. Critical data (e.g. ledger/account-state records) must retain its mandated 3-region durability floor at all times — a cost optimization must never silently weaken this disaster-recovery guarantee below policy. Freshness requirements genuinely differ by dataset class, and not all data requires full active replication to every region: archived and low-priority data may use slower, narrower replication where explicitly allowed. Bring cost back to a sustainable level without ever breaching the durability floor the business depends on.
- Difficulty
- Senior Staff
- Expected Time
- ~20 min
- Category
- Infrastructure & Capacity
Skills
- Distinguishing durability-floor policy from freshness/latency requirements
- Reasoning about per-data-class replication scope, mode, and cost as independent levers
- Recognizing a cost optimization that looks safe right now but breaches a durability guarantee
- Evaluating two genuinely different cost-reduction mechanisms (scope narrowing vs. mode/batching) on their own merits
System Architecture
- US Region
- EU Region
- APAC Region
- Replication Layer
- Primary Data Store
- Archive / Object Storage
This is a Pro scenario. Upgrade to Pro to unlock it.
Upgrade to ProBack to Incidents