If this is so and the thundering herd is the problem, then you can
easily verify it with modest web traffic. Start up apache with a large
number of processes (and fiddle with the parameters so they don't go
away). Then start a small amount of web traffic. If the thundering
herd is present, you should get cpu usage that scales with the square
of the number of processes -- N processes woken for every accept(),
which then walk down a run queue of length N.
Scale N up until you see the problem.
-- g
-
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/