Here's a direct link to the pdf[1] because the socallinuxexpo site returns a 404.<p>[1] <a href="http://www.brendangregg.com/Slides/SCaLE_Linux_Performance2013.pdf" rel="nofollow">http://www.brendangregg.com/Slides/SCaLE_Linux_Performance20...</a>
This makes me want to run to Solaris. How disgustingly fragmented, mostly unusable piece of crap Linux performance and profiling tools are. As someone who has seem even top break due to kernel changes, seen perf segfault time to time and seen that the dtrace ports (or SystemTap for that matter) most of the times lock up/panic the system - Solaris (or any other modern UNIX variant like HP-UX) feel so much more reliable, cohesive and just a better overall experience in profiling and performance analysis departments.<p>Well at least perf works most of the time on Linux.
Couldn't have asked for better timing. I'm currently debugging performance issues on a production machine and the summary of tools on slide 16 is really useful. I like the breakdown into beginner, intermediate and advanced.<p>For anyone skimming, be sure to check out Methodologies (slide 86), which covers 4 different approaches to performance analysis.
full blog post with video (was on HN already once or twice) <a href="http://www.joyent.com/blog/linux-performance-analysis-and-tools-brendan-gregg-s-talk-at-scale-11x" rel="nofollow">http://www.joyent.com/blog/linux-performance-analysis-and-to...</a>
Video available online as well.
<a href="https://www.socallinuxexpo.org/scale11x/presentations/linux-performance-analysis-and-tools" rel="nofollow">https://www.socallinuxexpo.org/scale11x/presentations/linux-...</a>