Reading latency through dependency graphs
Why the loudest service is rarely the whole story.
Blog
Notes from the tracing studio
Field writing on dependency graphs, sampling trade-offs, and how teams talk about latency.
Correlating traces across service boundaries
Headers, baggage, and the gaps between “instrumented” and “useful.”
When fan-out hides the real bottleneck
Parallelism that looks green until a single waiting edge dominates.
Sampling strategies for high-throughput meshes
Budgets that protect rare edges without drowning storage.