On Tue, 2006-11-21 at 16:11 +0300, Dmitry Butskoy wrote: > Arjan van de Ven wrote: > > >>Dave, > >>As you have appeared here, could you please comment why HZ was changed? > >> > >>I have some confusion on this. > >>First, the upstream kernel has this changed (since 2.6.13 ?) in backward > >>direction, i.e. from 1000 to 250. > >> > >> > > > >actually the upstream kernel changed to offer a choice, with 3 options, > >1000, 250 and 100. > > > Sure! But I meant the *upstream default" was chosen to be 250 (not 1000). upstream "defaults" don't mean much; Fedora ships a kernel config that deviates from that in hundreds and hundreds of places. > >does it matter? This is only internally visible to the kernel, not to > >userspace > > > > > But these changes are appreciable enough. For production servers (more > system CPU time), this is probably so little it's not even measurable. > for laptops (less battery life), etc. actually the "knee" in battery life for laptops is between 100 and 250, not between 250 and 1000. Both 250 and 1000 have very similar power use; only when you go really below 250 are you going to get some extra savings. (And only if you don't use USB at all.. which Fedora by default enables) > And note the ntp > issue discussed here too... .... with the upside that userspace can now sleep far more accurately which helps media playback, the desktop and all kinds of business applications. In addition the scheduler now has better information and makes better decisions as a result. Even ntp should be better off with this more finegrained data, but I can believe it'll need to get used to it for some time ;) -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
