News for hackers. Stuff that matters.
Re:Broken Performance Tools [pdf] by brendangregg on Tuesday November 24, 2015 @12:50PM (#10622258)
In case hackernews traffic brings down my website: http://www.slideshare.net/brendangregg/qcon-2015-broken-perf...
Re:Broken Performance Tools [pdf] by ElijahLynn on Tuesday November 24, 2015 @08:18AM (#10620482)
Where is the video of this talk? Slides are only half the story.
Re:Broken Performance Tools [pdf] by ElijahLynn on Tuesday November 24, 2015 @08:20AM (#10620487)
Re:Broken Performance Tools [pdf] by phaas on Tuesday November 24, 2015 @10:34AM (#10621283)
The video of the presentation is currently available for qcon sf attendees only: http://www.infoq.com/conferences/qconsf2015

The videos are released publicly over time...

Re:Broken Performance Tools [pdf] by moconnor on Tuesday November 24, 2015 @03:40AM (#10619769)
One of the criticisms of "top" is that it clears the screen and you cannot turn this off. "top -b" works fairly well for this.
Re:Broken Performance Tools [pdf] by brendangregg on Tuesday November 24, 2015 @12:55PM (#10622293)
ah, thanks, I forgot about this. Although it doesn't quite do what I want: I want it to truncate to the screen (like normal top), and just print the next summary without clearing the screen. That way you'd use it like normal top, but in case you wanted scroll back it'd be there.
Re:Broken Performance Tools [pdf] by dhimes on Tuesday November 24, 2015 @10:13AM (#10621122)
I just run it in its own terminal.
Re:Broken Performance Tools [pdf] by Ono-Sendai on Tuesday November 24, 2015 @08:20AM (#10620489)
You do need to know what you are doing when doing benchmarking. Doesn't mean that ~100% of benchmarks are wrong.
Re:Broken Performance Tools [pdf] by feld on Tuesday November 24, 2015 @01:11PM (#10622382)
They're almost always not able to be replicated, and measuring something almost always skews the results!

Check out gnn's talk on this tooled towards networking in FreeBSD

https://www.youtube.com/watch?v=9BcdRHNTdf4