[ale] cpu load 8.5 on usb to usb move

[Michael B. Trausch]

On Fri, 2011-02-25 at 15:25 -0500, Geoffrey Myers wrote:
> I understand that.  Why is it that the load average is so high, yet
> the 'mv' process shows 8-11% cpu? 

Because you're waiting on I/O, and multiple kernel threads are likely
blocked in the process.

IOW, it's not mv itself that is spiking the load average.  It's the I/O
through the USB ports.

As to the exact method that is used to determine how high to spike the
load when the system is waiting heavily on I/O, I don't know much about
that.  The only way to answer that, I think, would be to dig into the
source code.

My guess is that there are multiple kernel threads that are actively
involved in the management of the communications over the USB bus, and
they are probably blocked on each other in such a situation.  If the two
ports are on two separate USB controllers, that means that you're
probably talking about (at least) two kernel threads and the "mv"
process all in the D state at the same time.  And that will bring the
load average through the roof.

	--- Mike