Some API requests traverse many microservices and take unusually long. How do you determine which service contributes the most latency for those requests?

Stackdriver Trace (now Cloud Trace) is designed for distributed tracing, allowing you to visualize the end-to-end path of a request through multiple microservices. By instrumenting your application, Trace automatically collects and displays latency data for each service involved in a request, pinpointing bottlenecks. Setting timeouts might fail requests faster but doesn't identify the source of latency. Sending custom metrics to Monitoring could show overall latency but wouldn't break it down per service without significant custom effort. Looking for insights in Monitoring might indicate high latency periods but won't inherently reveal which specific service is the culprit in a multi-service request.