Mark,

> One other "symptom."  When I "echo 1 > hz_timer", the status of
> ksoftirqd_CPU0 becomes "SWN."  When I "echo 0 > hz_timer" it becomes "RWN."

Yes, that's the problem.  As I recall, the issue is that even though
ksoftirqd_CPU0 is really active only very rarely, for some reason the
timer interrupts 'conspire' to arrive usually just during those short
periods, so that to the stochastic load computation, this process
appears to be always running ...

This phenomenon appears to be caused by some interaction between the
delivery of I/O interrupts and timer interrupts when running under
VM (or LPAR with shared CPUs, for that matter).  We haven't quite
understood this whole issue yet, unfortunately.  (Maybe Martin knows
more details ...)

Bye,
Ulrich

--
  Dr. Ulrich Weigand
  [EMAIL PROTECTED]

Reply via email to