A vendor-neutral standard for traces, metrics and logs, so instrumentation outlives the monitoring contract.
What we reach for first
Every system we build or take over. It goes in before optimisation, which is what makes an improvement attributable. It also makes the monitoring vendor a decision you can reverse, because the instrumentation is written once and belongs to you rather than to the contract.
Nothing — though tracing everything at full sampling on a high-volume service is a bill, and the sampling policy is a decision, not a default. Traces on everything at full sampling is an expensive way to store noise; retention and sampling are numbers somebody signs off, not defaults nobody read.
Trace context propagated across service boundaries, and a baseline captured before any performance work begins. One trace identifier follows a request across every service it touches, which turns “the system is slow” into a named span in a named service — the difference between an investigation and a discussion.
Where this appears on the site
Derived from the stacks published on those pages, not written here — so this list cannot claim something the page it points at does not say.
Also in delivery and observability
Tell us what it is running, what it costs you today, and what you need it to do next. A senior engineer will tell you what we would keep and what we would change.
