Using dotnet-traceĭotnet-trace allows us to enable the runtime CPU sampling provider ( Microsoft-DotNETCore-SampleProfiler). When combined with CPU sampling, it is the best non-invasive profiling solution. It uses the system scheduler/dispatcher events to get detailed information about application CPU time. For such scenarios, we should rely on Thread Time profiling.

It won't work for analyzing waits in the applications. Thus, this technique is excellent when diagnosing high CPU usage of an application. CPU sampling is about collecting samples in intervals: each CPU sample contains an instruction pointer to the currently executing code. We could either use CPU sampling or Thread Time profiling.

Pidstat -p ALL 1 Collecting process traces # show CPU usage for all the processes (-p ALL) and refresh every second # filter processes with pgrep and show their info with ps # sort all processes (-sort) by their CPU usage (from highest to lowest)

# show processes whose effective user ID (-u) is root and select only specific columns (-o) # show all processes (-e) running in the system with some details (-F)

Process monitor windows cpu usage