bk99.de entertain the web since 1997

bk99.de
HN · PerformanceCurated

Narrowing down Linux performance in sixty seconds

Summary

Brendan Gregg arranges ten standard commands into a quick procedure for load, CPU, memory, disks and network. The procedure comprises ten commands. The first investigation takes about sixty seconds.

Ideas

  • uptime shows the development of load over several time windows.
  • dmesg reveals kernel and hardware events.
  • vmstat combines processes, memory, paging and CPU time.
  • iostat makes disk utilisation and queues visible.

Insights

  • A fixed diagnostic procedure prevents premature single explanations.
  • System metrics only gain meaning when viewed together over time.
  • Quick triage decides which deeper tools make sense afterwards.

Facts

  • The examples use common Linux tools.

Recommendations

  • Collect timestamped output before making changes.
  • Only dig deeper into measured bottlenecks with specialised tools.

References

Read the original article

Search the Web Archive

HN · 2022

How fast are Linux pipes?

Linux pipes were made considerably faster through larger buffers, fewer copies and optimised use of pages.

HN · 2019

Booting the Linux kernel faster

Feng Tang shows measurement methods and parallelisation that cut the kernel start of a virtual platform from three seconds to 300 milliseconds. The presentation is from the Linux Plumbers Conference 2019. The measured kernel time fell from 3,000 to 300 milliseconds.

HN · 2017

The mystery of Linux load averages

Brendan Gregg reconstructs why Linux includes not only runnable processes but also tasks in uninterruptible wait in the load average. Linux shows load averages for one, five and fifteen minutes. The relevant patch dates from 1993.