Why this cost is so easy to miss at design time
Compute and storage costs are intuitive and visible on every pricing calculator; egress and cross-region transfer costs are buried in a separate pricing page most architects don't open until the bill arrives. A design that replicates data across regions for resilience can multiply this cost without anyone noticing until it's load-tested.
The fix isn't avoiding multi-region architecture — resilience often genuinely requires it — it's pricing the transfer explicitly during design, the same way compute is priced.
Patterns that reduce it without sacrificing resilience
Keeping chatty, high-volume service-to-service calls within a single region, and reserving genuine cross-region calls for asynchronous, batched, or infrequent replication, preserves most of the resilience benefit while avoiding the worst of the transfer cost multiplier.
The architecture diagram doesn't show a cost. The invoice always does. Price the transfer at design time, not at scale.MNB Cloud Practice
Key actions
- Price cross-region transfer explicitly during architecture design, not after launch
- Keep chatty service-to-service calls within a single region where possible
- Reserve cross-region calls for async, batched, or infrequent replication
- Load-test cost projections, not just latency, before go-live