Profiles, the Missing Pillar: Continuous Profiling in Practice

With Continuous Profiling (CP) you capture resource usage (such as CPU, memory, I/O, etc.) over time, enabling you to pinpoint the (source) code that is slow or causes an issue. In recent times, CP has become mainstream and a number of open source projects such as Parca, Pyroscope, or CNCF Pixie are now available to benefit from CP in your own environment. We will have a look at the origins and the underlying motivation of CP and discuss cost and benefits of using CP in production. Finally, we will make the case that profiles are, in addition to logs, metrics, and traces, the missing pillar of observability.