CODEWITHSUNDEEP
performanceprofilingmonitoringbenchmarkingperformance-testing
jimakos17
jimakos17

Reputation: 935

How dstat measure latency?

dstat provides --top-latency and --top-latency-avg plugins in order to show processes with highest latencies (in ms).

How does it work this? I run a simple dd, which output it's latency, run dstat with these plugins, I sum the latencies for this process and the result wasn't the same.

Actually, it was higher! dd displayed 13s and dstat 21s, after summing the milliseconds.

Do I have understand it correctly? Is there any other tool which measure the latency per process?

Upvotes: 1

Views: 787

Answers (2)

Mircea Vutcovici
Mircea Vutcovici

Reputation: 2404

The dstat command is using information from /proc/<PID>/schedstat. the format of this file is documented at https://www.kernel.org/doc/Documentation/scheduler/sched-stats.txt

You can see how it is used in the dstat's source code: https://github.com/dagwieers/dstat/blob/master/plugins/dstat_top_latency.py#L42

Upvotes: 1

justpusher
justpusher

Reputation: 41

Consider using the latencytop tool

Synopsis latencytop [--unknown] [processes...]

Description (from the manual):

"Latencytop is a Linux* tool for software developers (both kernel and userspace), aimed at identifying where in the system latency is happening, and what kind of operation/action is causing the latency to happen so that the code can be changed to avoid the worst latency hiccups."

Upvotes: 1

Related Questions