Engineering 4 min read
LatAm Market: How We Cut p95 Latency to 44ms
A new edge region in São Paulo brings sub-50ms API latency to Latin American operators. Here's the architecture behind it.
VI
Alex Rivera
VP Infrastructure · May 20, 2025
Edge deployment: sa-east-1
Our new São Paulo edge region is live. LatAm operators now see 44ms p95 latency — down from 180ms+ when traffic routed through us-east-1.
Architecture
- 9 edge regions globally: Dublin, Frankfurt, Virginia, Oregon, São Paulo, Mumbai, Singapore, Tokyo, Cape Town
- Automatic failover: if an edge degrades, traffic reroutes to the nearest healthy region
- Read replicas at every edge for allowlist lookups (sub-5ms)
- Write-through to the primary for usage events (eventual consistency <2s)
The latency map
Our public API latency map now shows all 9 regions with live p50/p95 numbers. Operators can verify their closest edge before integrating.
Why LatAm matters
LatAm is our fastest-growing region — 41% MAU growth QoQ. Sub-50ms latency is table stakes for retention; at 100ms+, players churn.
Latency numbers are placeholder targets. Wire to real edge telemetry on launch.
#Latency#Edge#Infrastructure#LatAm