numbers don't lie ...
Dmitry Morozovsky
marck at rinet.ru
Mon Sep 25 22:48:48 PDT 2006
On Mon, 25 Sep 2006, Dag-Erling Sm?rgrav wrote:
DS> > > > My experiments show that if you have enough memory to host radmdrive for
DS> > > > /usr/src you'd better leave it for caching - there were no statistically
DS> > > > meaningful performance difference, at least on machines with 1G+ RAM.
DS> > > Really? My measurements show the opposite (on a system with 16GB of
DS> > > RAM).
DS> > My last test on amd64/dualcore with 4G of RAM and -j4 shows
DS> > (buildworld+buildkernel):
DS> >
DS> > ==> /tmp/buildlog <==
DS> > 1996.45 real 3032.94 user 624.83 sys
DS> > Script done on Tue Sep 19 14:44:54 2006
DS> >
DS> > ==> /tmp/buildlog.md <==
DS> > 1957.45 real 3033.93 user 585.78 sys
DS> > Script done on Tue Sep 19 15:20:42 2006
DS> >
DS> > Second one was with 512M/4k/512 swap-backed md, the former with /usr/src on the
DS> > gmirror'ed pair of SATAs.
DS>
DS> Seems to me that your own numbers contradict you. You saved about 40
DS> seconds (2%) by keeping /usr/src in a ram disk.
Well, maybe I did used wrong wording (2% is statistically meaningful), but for
me 2% is not worth all the efforts needed to organize ram disk
infrastructure...
Sincerely,
D.Marck [DM5020, MCK-RIPE, DM3-RIPN]
------------------------------------------------------------------------
*** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck at rinet.ru ***
------------------------------------------------------------------------
More information about the freebsd-hackers
mailing list