Skip to main content

Profiling 101

Understand basics of CPU profiling in 5 minutes ⏱️!

Example​

This guide will demonstrate CPU profiling using Go, but these fundamentals apply to any language. Let's walk through an example of CPU profiling.

Capturing data​

Take the following example of a Go program that has a main function, that first calls iterateLong which calls iterate with 9 billion iterations, and then iterateShort, which calls iterate with 1 billion iterations.

package main

func main() {
    iterateLong()
    iterateShort()
}

func iterateLong() {
    iterate(9_000_000_000)
}

func iterateShort() {
    iterate(1_000_000_000)
}

func iterate(iterations int) {
    for i := 0; i < iterations; i++ {
    }
}

Note: Save this snippet to iterate.go and run it with time go run iterate.go.

When executed this program takes 5 seconds to execute in total (on an AMD Ryzen 5 3400GE CPU). With profiling we can understand what was executing during those 5 seconds and for how long. For the sake of simplicity, a sampling CPU profiler looks at the "current" stack trace 100x per second (the sampling rate is typically configurable, but 100x is both common and easier to calculate with).

Data format​

With a profiler running during the execution of the above program it records a profile, that produces the following data in folded stack trace format:

iterate;iterateLong;main    450
iterate;iterateShort;main   50

Parca uses the open standard pprof, which is optimized to use as little space as possible, but folded stack traces are great for human readability.

10% (50 samples observed out of 500) of the time was spent in the iterate function called by iterateShort and 90% (450 samples observed out of 500) of the time was spent in the iterate function called by iterateLong.

Visualizing​

Using this data, a popular way to visualize profiling data is using flamegraphs.
If they are built with the root at the top: Flame Graphs.

Profiling 101 Flame Graph

Recap​

In this guide you have learned the fundamentals of CPU profiling:

  1. How data is captured: by observing the executed stack traces 100x per second.
  2. What the raw data looks like: folded stack traces, and the optimized pprof format.
  3. Useful ways to visualize data: Flame Graphs.

Congrats and happy profiling! 🎉