Bet 14 · Modern cloud-native performance
Performance
wrldOS runs on a multi-region active-active stack. Page paints, API responses, EHR reads, and real-time event delivery are measured continuously and shown live below. This is the architectural difference from legacy EDC platforms — sub-second is the target, not the exception.
Live · all systems normal
Page first paint
312ms
p50 298ms
p95 412ms
p99 488ms
Target ≤ 400ms · alert ≥ 1.50s · sampled every 2s
API response (p50)
47ms
p50 47ms
p95 92ms
p99 138ms
Target ≤ 100ms · alert ≥ 400ms · sampled every 2s
Real-time event delivery
84ms
p50 78ms
p95 142ms
p99 198ms
Target ≤ 200ms · alert ≥ 800ms · sampled every 2s
EHR FHIR read
211ms
p50 198ms
p95 412ms
p99 612ms
Target ≤ 500ms · alert ≥ 1.50s · sampled every 2s
Regional uptime — 30 days
Active-active across 4 regions
| Region | Location | 30-day uptime | Status |
|---|---|---|---|
| us-east-1 | Virginia | 99.987% | HEALTHY |
| us-west-2 | Oregon | 99.992% | HEALTHY |
| eu-west-1 | Ireland | 99.963% | HEALTHY |
| ap-southeast-1 | Singapore | 99.971% | HEALTHY |
Architectural posture
Multi-region active-active
wrldOS runs in 4 regions simultaneously. Failover is sub-second; no single region is a single point of failure.
Push, not pull
Every event — query, deviation, lab, signature — is pushed to the client over a persistent WebSocket. No 30-second polling intervals.
Edge-cached protocol intelligence
Protocol Q&A, semantic search, and cross-document reasoning run at the CDN edge. Most answers return inside 200ms.
Stateless workers, hot autoscale
Every API worker is stateless. Traffic spikes (sponsor demos, audit exports, end-of-study lock) scale horizontally in 8 seconds.
Configuration and credentials are managed in Settings.