Rob van der Heij wrote:
On Fri, May 1, 2009 at 1:12 AM, Leland Lucius <[email protected]> wrote:
But, I'm curious why hottplugging CPUs doesn't do WAS any good?
I was hoping you would ask ;-)
I've been called a straight man before. :-)
The problem with idle Linux guests with multiple virtual CPUs is that
Linux spreads the little work it has over all these virtual CPUs. So
even when virtual CPUs properly drop from queue, CP will rarely find
all virtual CPUs idle at the same time. This means that memory
management on z/VM gets frustrated. But when cpuplugd "unplugs"
additional virtual CPUs when the server is idle, CP has chance to take
resources away.
Gotcha.
WAS (and JVM) have various things running that poll for work and thus
never drop from queue. They can't be helped by by unplugging CPUs
because the remaining one will be in-queue all day as well.
But, if there's more VCPUs lying about, doesn't z/VM have to schedule
them? If they're "offline" in the guest, doesn't z/VM see that and
ignore 'em when figuring out how to schedule stuff? (I'm trying to talk
WAY over my expertise here, so ignore me if you like. ;-))
Also, if load averages increase sufficiently, doesn't adding another
VCPU help smooth out the load (mainly during spikes)?
Leland
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
