# Rebuild one request across edge, backend, and dependencies.

> TrafficWar reconstructs distributed request waterfalls from shared trace IDs across HTTP, database, Redis, and S3 spans.

Send related spans with one trace_id and label. TrafficWar groups them into a request waterfall and separates inclusive span duration from time spent in lower tiers.

## A small, explicit trace contract

Tracing uses the same event ingestion path as every other TrafficWar event. No separate collector is required: add trace_id, span_kind, source, operation_type, and latency_ms to the events you already capture.

## Tier-aware waterfall reconstruction

- Client HTTP spans represent the edge-facing request.
- Server HTTP spans represent backend execution.
- Database, Redis, and S3 client spans represent dependencies.
- Repeated dependency calls are retained and summarized in the journey.

## Trace identity is not metric cardinality

TrafficWar uses trace_id for request reconstruction and search. It does not place trace IDs on metric aggregation keys, which keeps per-route metrics bounded independently from trace volume.

## Related

- [Connect spans with one trace ID and route label.](https://trafficwar.tech/docs/traces.md)
- [Watch requests move through every application tier.](https://trafficwar.tech/features/live-map.md)
- [Capture your first TrafficWar event with Node.js.](https://trafficwar.tech/docs/getting-started.md)