On Fri, May 8, 2009 at 1:25 PM, Harder, Pieter <[email protected]> wrote:
> And now for the really stupid question: does enabling cpuplugd with CPU_MIN=1 > actually gain something? I have turned it on for one of my SAP dyadics and my > idle level seems to have gone up instead of down. I now see cpu spikes at the > cpuplugd UPDATE interval that aren't there with cpuplugd disabled. Obviously > running cpuplugd costs something and falls into the chapter of disabling > unneeded services. This only makes sense when the processing saved in CP > amounts to more than the cost of running cpuplugd. I see more stupid answers than stupid questions, but that may be because I often give the answers ;-) When switching off the excessive virtual CPUs makes the difference to drop from queue, it may be well worth the investment of some extra CPU usage to manage that. So you pay some CPU to save storage. If it does not save storage, you might not want to spend CPU on it. The other motivations for it don't apply well to Linux on z/VM. Something like cpuplugd is an ugly hack. I wish Linux would not engage all virtual CPUs when the workload does not take advantage from it. Simply loading an enabled wait PSW would be enough. The idea behind CPU affinity that causes this does not apply to Linux on z/VM as I see it. Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ ---------------------------------------------------------------------- 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
