Trace, Profiler, Error Reporting, and Managed Prometheus

Speaker notes

• Cloud Trace shows where time is spent across a request's full path through multiple services — the concrete tool behind 'the database query specifically took 800ms of a 900ms total,' a materially more actionable diagnosis than 'the request was slow.'
• Cloud Profiler continuously samples CPU and memory usage in production at low overhead, for free — it answers a different question than Trace: where does the application spend its OWN CPU/memory, useful for finding an inefficient function even when no single request looks slow in a trace.
• Error Reporting automatically groups and surfaces exceptions from structured logs, also free, with no separate configuration beyond logging exceptions in a recognized format. Google Cloud Managed Service for Prometheus gives the same PromQL query language and scrape model as self-hosted Prometheus, without operating the collection infrastructure — genuinely useful for a team already fluent in Prometheus from another environment.

Deck map