Comment 19 for bug 1833322
Revision history for this message
| Christian Ehrhardt (paelzer) wrote Re: Consider removing irqbalance from default install on desktop images | #19 |
© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
bbfa235
(Get the code!)
Hi Mike
> SUSE ... says that the first step to get there is to disable irqbalance
I've read the same, IMHO that is just "if you want to manually tune, disable
it" which does not imply that it is bad to have it. But this is how I read
it, I have not talked to the authors to get their underlaying reasoning.
> Applications vendors ... currently recommend removing irqbalance
The only one that does so AFAICS is cpufreq and everyone else just links
to their reasoning and follows. And even some statements there like
"If you are still running irqbalance, you are not getting the maximum
performance your system is capable of!" are hard to believe as a general
statement - especially without data across a wide variety of system types
and workload.
As we have seen as well in the references linked, irqbalance helps just as
much for "maximum performance" in many other cases.
> I found this blog (https:/
Thanks, every extra background we find will only help (except for those
joining later to read more).
> The question I have is, if Ubuntu is Debian Branch, and we long ago went
> from having different kernels for desktop & server in ubuntu-base, but do
> have ubuntu-server packages and ubuntu-desktop packages, where things could
> be different, why is this still a broad sweep as a default install "for all"?
Because there was no well-funded conclusion like "it really is bad for
environment X" to remove it. You are right that there are no technical blockers
to make it e.g. kept in servers but no more the default in Desktop.
After all it is already dropped in cloud-images used in virtual environemnts as
it had a more clear reasoning and argument there.
And there are also cases where irqbalance missing caused performance impact
and bug reports like the already mentioned [1] (clearly high scale server
though)
> I am happy that this is getting discussed properly now so that we can
> relook at this, and what it means to us today.
Ack, that is why I tried to compile all I've found into one place.
[1]: https:/