From Traces to Topology
Every span records who called whom. Aggregate millions of these and you can derive a service dependency map, a graph where nodes are services and edges are calls between them.
What the Edges Tell You
- Direction: which service depends on which.
- Volume: how many calls flow along an edge.
- Health: error rate and latency per edge, not just per service.
This map is built from real traffic, so it reflects what actually happens rather than an outdated architecture diagram.
Why It Matters
The map answers questions a static diagram cannot. When payments slows down, you can see every service that calls it and gauge the blast radius. You can spot a surprise dependency nobody documented, or a service that should be deprecated but still receives traffic.
It is also the foundation for the next step, finding the critical path through a single request.
Key idea
A service dependency map is a graph derived from real trace data showing which services call which, with volume and health on each edge.