Feature

Profiling

Continuous profiling: CPU and memory flame graphs in production, low overhead.

Continuous profiling captures CPU and memory-allocation flame graphs straight from production processes with minimal overhead. You see which function burns CPU and where memory leaks — without reproducing it on a bench. Comparing before/after-release profiles surfaces a regression down to a specific stack frame.

Key properties

  • CPU and allocation flame graphs from production
  • Low overhead — safe to keep always on
  • Before/after-release profile diff by stack frame
  • Tied to service and instance via shared labels
  • Drill from an expensive endpoint in a trace to its profile

Related features