I seem to have done an unlucky division of my Linux Expo papers
between my briefcase and larger luggage -- one interesting part of
them didn't follow me thru the JFK airport; they should follow me
tomorrow (or so the airline told me..) So from my non-reliable memory...
At Linux Expo there was a Work-In-Progress report about Alpha Iprobe
performance analysis tool, and its usage results at a web-bench when
system is running Apache.
With it, it is clear that there "Thundering Herd" happens, and that
the scheduler is used heavily because it is doing linear scans of
process descriptor lists.
Same engineers also developed a new constant time scheduler which
they intent to submit to 2.3. (The change is "too large" for it to
be accepted into 2.2.)
With "Wake One" and the scheduler changes, Iprobe does show that the
time spent in small number of scheduler addresses does down dramatically.
(From 30% to 0.1% of realtime, or some such...)
> -- mingo
/Matti Aarnio <matti.aarnio@sonera.fi>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/