“The sampling was at 99 Hertz across all CPUs, to minimize overhead (instead of, say, 1000 Hz), and to avoid lockstep (with any power-of-10 Hz task).”

Just curious, are there many tasks that run on a power-of-10 schedule? Also, what kinds of tasks do you need to avoid colliding with (whether power-of-10 or not). Any examples? :)

I have been using “fishbone” methodology in my troubleshooting efforts in enterprise environments. I am now going to leverage both (Fishbone and USE) and I am sure, sysadmins will be thrilled to see the quick results that this tool will deliver – particularly when users are screaming of “slowness in the system”.

Appreciate your breakdown of key metrics.

-Harsha Nippani

I made a similar visualization attempt a while back, http://wingolog.org/archives/2009/02/09/visualizing-statistical-profiles-with-chartprof ; but yours wins for two reasons:

* Hot functions (wide bars) will have space enough for the name of the function — they are oriented in the direction of text
* Mouseover allows you to present all the informaiton without having to allocate pixels to all of it.

Again, very nice work. It would be nice to assign significance to color: e.g. the more time spent in a function (as opposed to in its callees) would make it redder.

Thanks for sharing!