The platform is built to scale horizontally and meet interactive performance targets. The figures below are the project's stated targets (and reported current values) — validate them against your own environment.
| Metric | Target | Reported |
|---|---|---|
| Page load | < 2s | ~1.5s |
| API p99 latency | < 200ms | ~150ms |
| Login | < 1s | ~500ms |
| Service startup | < 30s | ~20–25s |
Scaling approach
Services are stateless and use stateless JWTs, so there is no session affinity — capacity is added by running more instances, discovered through Eureka and balanced by Spring Cloud LoadBalancer (round-robin) and Nginx (least_conn). The infra service can add or remove replicas at runtime; see Service Deployer.
Database
A single PostgreSQL instance with HikariCP pooling (max pool 20, min idle 5). Production deployments add streaming replication (primary for writes, replica for reads) with failover.
Caching
- Service layer:
@Cacheable/@CacheEvictwith a configurable TTL (≈1 hour). - Frontend: TanStack Query for request dedup, stale-while-revalidate and optimistic updates.
- Nginx: long
max-agefor versioned assets, no-cache forindex.html, gzip enabled.