mfs and buildworlds on the SunFire x4600
Mars G. Miro
marsgmiro at gmail.com
Tue May 8 03:45:19 UTC 2007
On 5/8/07, Oliver Fromme <olli at lurza.secnetix.de> wrote:
> Mars G. Miro wrote:
> > Oliver Fromme wrote:
> > > By the way, what are you actually trying to do? What is
> > > your goal? Do you need to reduce the buildworld time?
> > as i've mentioned in my original email, does mfs speed up I/O stuff ?
> Sometimes it does. But most of the time, a real disk
> partition with soft-updates on it is just as fast.
> With soft-updates, writing is asynchronous, i.e. it
> goes to RAM first, just like a memory disk. The data
> is later committed to disk in the background, so the
> processes don't have to wait for it. And once the
> data is in the cache, reading is just as fast (or even
> faster) as a memory disk. Note that /usr/src will
> fit in the cache easily if you have several GB of RAM.
> I usually have a memory disk as /tmp, but that's really
> just for historical reasons. And it's easier to clean
> up -- just umount it. ;-)
> > there's been a lot of threads in teh past that a buildworld on mfs
> > increases speed --- tho it might not be the appropriate test for
> > high-end machines (speaking of w/c I just gots a T2000).
> It depends on what exactly you want to test, and for
> what reason. You probably have already wasted much
> more time with your experiments and testing than you
> can ever save by using mfs for buildworld.
wasted my time? dont think so.
now we know buildworld on mfs dont really matter on high-end machines,
and it didnt even then when i tried it on my single-proc Opteron w/ 1G
of RAM almost 2 years ago on 5.X (i recall having to crash when i used
malloc but then this is documented)
and I'd think testing it on something like my x4100 w/ 8G of RAM may
produce the same results..
so teh conclusion would be, buildworld isnt teh appropriate test if
mfs does really speed things up, other apps/tools may be much more
appropriate --- that or, does mfs speeding things up really work?
remains to be seen ...
> > there's prolly other appropriate apps/tools for mfs-testing ...
> I don't think it makes much sense to benchmark mfs.
> It is a known fact that a real tmpfs (like Solaris and
> Linux have) would be better. I think it's even listed
> on the FreeBSD ideas web page, but nobody is actively
> working on it, AFAIK. On the other hand, I'm not 100%
> convinced that it would be worth the effort either.
it does to me, however, and perhaps other people too ;-)
> It would be interesting to see how ZFS on a swap-backed
> vnode device would perform on FreeBSD 7-current (with
> and without compression).
> Best regards
> Oliver Fromme, secnetix GmbH & Co. KG, Marktplatz 29, 85567 Grafing b. M.
> Handelsregister: Registergericht Muenchen, HRA 74606, Geschäftsfuehrung:
> secnetix Verwaltungsgesellsch. mbH, Handelsregister: Registergericht Mün-
> chen, HRB 125758, Geschäftsführer: Maik Bachmann, Olaf Erb, Ralf Gebhart
> FreeBSD-Dienstleistungen, -Produkte und mehr: http://www.secnetix.de/bsd
> One Unix to rule them all, One Resolver to find them,
> One IP to bring them all and in the zone to bind them.
More information about the freebsd-stable