On Fri, Jul 18, 2008, Peter Jeremy wrote: > On 2008-Jul-18 08:37:25 +0200, Alexander Leidinger <Alexander_at_Leidinger.net> wrote: > >Are you aware that the parallel starting in Solaris 10 reduced the > >booting time by a nice percentage? > > Given that Solaris boots in geologic time, this probably wouldn't > be difficult. Solaris actually boots significantly faster than FreeBSD on my desktops, but that's not because of the parallel rc startup--- it's mainly for the completely orthogonal reason that the USB stack and the ata(4) driver in FreeBSD take an eternity to probe. So I suspect that people interested in making FreeBSD boot faster could find lots of low-hanging fruit even without touching rc. > Parallel starting is not guaranteed to be an improvement. Starting a > whole pile of processes that are I/O bound during initialisation > (think squid or some databases) may be worse than starting them one > at a time. Likewise, a whole pile of processes that are CPU bound > will just thrash the scheduler. (Though parallel starting of I/O and > CPU bound processes should be a win). Parallel service startup in Solaris does seem to save time (although this is on a dual core; I haven't tried a UP system). Keep in mind also that many rc scripts spend a few seconds waiting on the network, and that time could be better spent starting something else. That said, parallel startup doesn't make an enormous difference, and that wasn't the main reason Sun wrote SMF. The main motivation was to be able to automatically restart failed services and their dependencies, and provide better administrative tools. If you can find a way to decrease the boot time considerably, many people won't care, but I'm sure some will greatly appreciate it. There are two ways to cut downtime in half: make the system crash half as often, or make it recover twice as fast. The latter is probably an easier task. :)Received on Fri Jul 18 2008 - 12:25:46 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:33 UTC