Reading latency through dependency graphs

When a request slows down, most teams open a trace waterfall and stare at the tallest bar. That bar is rarely the full sentence. Application analytics for microservice dependency tracing asks a different first question: which edges on the path changed shape?

Graphs before waterfalls

A dependency graph compresses ownership, direction, and multiplicity. You see that checkout fans into inventory, pricing, and a recommendation sidecar — and that the sidecar is the only edge that cold-starts after a quiet hour. The waterfall still matters for timing detail, but the graph frames the story you will tell in the incident channel.

Name the waiting edges

Not every long span is work. Some are waiting. We teach learners to annotate waiting edges explicitly so product partners stop funding “optimizations” that cannot shrink queue time. This habit alone changes which tickets get prioritized.

Practice the narration

In Service Graph Observatory we rehearse a sixty-second explanation: journey, critical edges, what changed, what we will verify next. The goal is not eloquence; it is shared orientation. Latency becomes a map you can walk together.

Explore the flagship course if you want guided practice with your own anonymized graphs.